Introduction
The triangle singularity (TS), also known as Landau singularity, is a leading singularity of a one-loop triangle diagram. It can occur when the three internal particles simultaneously approach their mass shells in a pinched route and propagate along a classically allowed configuration [1–3]. Once the corresponding kinematic condition is satisfied, the loop integral can develop a logarithmic enhancement that may appear as a narrow peak in the invariant-mass spectrum of the two final-state particles, thereby mimicking the lineshape of a genuine resonance [4–9]. In the decay processes of hadrons there exist abundant open-channel thresholds, providing many kinematic windows for such configurations. Nowadays, more and more effects from the TS mechanism are investigated in theory and identified in experiment, which provide a special way for understanding the near-threshold dynamics in hadron-hadron scatterings [4, 6–13].
Over the past decade, the TS mechanism has been invoked in discussions of a broad range of experimental phenomena. The anomalously enhanced isospin-violating effect in the η ( 1405 / 1475 ) → 3 π \eta(1405/1475) \to3\pi η ( 1405/1475 ) → 3 π decay observed by BESIII has been attributed to the combined effect of a K ∗ K ˉ K K^*\bar{K}K K ∗ K ˉ K triangle loop and f 0 ( 980 ) f_0(980) f 0 ( 980 ) –a 0 ( 980 ) a_0(980) a 0 ( 980 ) mixing [4, 14–16]. The charged charmoniumlike state Z c ( 3900 ) Z_c(3900) Z c ( 3900 ) observed in Y ( 4260 ) → π + π − J / ψ Y(4260) \to\pi^+\pi^-J/\psi Y ( 4260 ) → π + π − J / ψ has also been examined from the perspective of a D 1 D ˉ ∗ D D_1\bar{D}^*D D 1 D ˉ ∗ D triangle mechanism [6, 7], and similar studies have been extended to a number of X Y Z XYZ X Y Z candidates [6, 7, 10, 17–25]. The hidden-charm pentaquark states P c P_c P c reported by LHCb in Λ b → J / ψ p K \Lambda_b \to J/\psi pK Λ b → J / ψ p K decays have likewise been examined within a triangle-mechanism interpretation [26–31]. A direct evidence for the TS mechanism is in connection with the axial-vector candidate a 1 ( 1420 ) a_1(1420) a 1 ( 1420 ) [8, 32]. Meanwhile, it shows that experimental data for the N ( 1700 ) N(1700) N ( 1700 ) region [33], the X ( 5568 ) X(5568) X ( 5568 ) [34], the Dalitz plot of A c + → K − π + π + A_c^+ \to K^-\pi^+\pi^+ A c + → K − π + π + [35], the X 0 ( 2900 ) X_0(2900) X 0 ( 2900 ) [36], etc. In most of these studies, it is generally emphasized that the satisfaction of the Landau condition can produce narrow peaking enhancement. But it should be recognized that in all these different cases the significance of the TS structure would also be affected by the dynamical details for the vertices. So far, systematic studies of the dynamical coupling effects with the TS mechanism have not yet been explored. This motivates us to investigate the manifestations of the TS mechanism when different coupling vertices are present.
For a fixed Coleman–Norton kinematic configuration [2], the detailed structure of the three vertices—including the partial wave carried by each coupling (S S S wave or P P P wave), the tensor structures associated with the vector or axial-vector intermediate states, and the Lorentz scalars generated in the amplitude decomposition, determines the relative weights of the three-point scalar integral C 0 C_0 C 0 and the two-point scalar integrals B 0 B_0 B 0 . When C 0 C_0 C 0 dominates, the invariant-mass spectrum generally exhibits a narrow TS peak. Comparable B 0 B_0 B 0 contributions can interfere with the triangle term and reduce the prominence of the peak. If the leading Landau logarithm is suppressed by the vertex structure, the remaining ordinary-threshold dependence and smooth background of the coherent amplitude can instead produce a cusp-like profile, without requiring an explicit B 0 B_0 B 0 term to dominate. To the best of our knowledge, a systematic comparison of how different dynamical couplings modify the TS lineshape for a fixed loop topology is still lacking. It is useful and necessary to investigate the diversity of the TS lineshapes from this perspective.
In this work, we systematically investigate the effect of the dynamical couplings at the triangle vertices on the TS lineshape. We consider ground-state 0 − 0^{-} 0 − or 1 − 1^{-} 1 − intermediate particles and allow the three vertices to couple through either an S S S or P P P wave, leading to twelve combinations of external and internal spin-parity quantum numbers. The corresponding effective Lagrangians cover seven types of vertices, namely P P S PPS P P S , V P P VPP V P P , V V P VVP V V P , V P A VPA V P A , V V V VVV V V V , V V A VVA V V A , and V V S VVS V V S [37 , 38 , 39 , 40 ] . With the aid of FeynCalc [41 , 42 , 43 , 44 ] , we use the Passarino–Veltman reduction [45 ] to express each loop amplitude as a linear combination of the scalar integrals A 0 A_{0} A 0 , B 0 B_{0} B 0 , and C 0 C_{0} C 0 . The Lorentz-scalar coefficients multiplying these integrals encode the dynamical information of the vertices under consideration. Comparing the relative sizes of these contributions near the kinematic position determined by the Landau equation allows us to identify the role of each scalar integral in shaping the spectrum. For the numerical comparison, we adopt the masses and kinematics of the η ( 1405 ) → π + π − π 0 \eta(1405) \to\pi^{+}\pi^{-}\pi^{0} η ( 1405 ) → π + π − π 0 system with a K ∗ K ˉ K K^{*}\bar{K}K K ∗ K ˉ K loop as a common benchmark for the twelve spin-parity assignments, using the f 0 ( 980 ) f_{0}(980) f 0 ( 980 ) parameters for the final-state resonance factor. This is the kinematic region where f 1 ( 1420 ) f_{1}(1420) f 1 ( 1420 ) and h 1 ( 1380 / 1423 ) h_{1}(1380/1423) h 1 ( 1380/1423 ) can have an S S S -wave coupling to the K ∗ K ˉ K^{*}\bar{K} K ∗ K ˉ channel. We find that, even for the same loop topology and kinematic condition, different vertex structures can generate three qualitatively distinct lineshapes.
The remainder of this paper is organized as follows. The formalism is introduced in Sec. II; the numerical results and their discussion are presented in Sec. III; and the main conclusions are summarized in Sec. IV.
The contribution of the triangle loop diagram for the 1 → 3 1 \to3 1 → 3 process is shown in Figure 1 . For a scalar triangle loop diagram, the amplitude is given by
Figure 1. Triangle loop diagram
i M = ∫ d 4 q 3 ( 2 π ) 4 1 ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) , q 1 = p 2 + q 3 , q 2 = p 3 − q 3 . (1) i\mathcal{M}=\int\frac{\mathrm{d}^{4}q_{3}}{(2\pi)^{4}}\frac{1}{(q_{1}^{2}-m_{1}^{2})(q_{2}^{2}-m_{2}^{2})(q_{3}^{2}-m_{3}^{2})},\qquad q_{1}=p_{2}+q_{3},q_{2}=p_{3}-q_{3}. \tag*{(1)} i M = ∫ ( 2 π ) 4 d 4 q 3 ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) 1 , q 1 = p 2 + q 3 , q 2 = p 3 − q 3 . (1)
Using the Feynman parameterization method and denoting the shifted loop momentum by q q q , this can be expressed as
i M = ∫ d 4 q ( 2 π ) 4 ∭ d x 1 d x 2 d x 3 2 δ ( ∑ i = 1 3 x i − 1 ) ( q 2 − D ) 3 , D = Y i j x i x j , Y i j = 1 2 [ m i 2 + m j 2 − Δ q i q j ] . (2) i\mathcal{M}=\int\frac{\mathrm{d}^{4}q}{(2\pi)^{4}}\iiint\mathrm{d}x_{1}\mathrm{d}x_{2}\mathrm{d}x_{3}\frac{2\delta\left(\sum_{i=1}^{3}x_{i}-1\right)}{(q^{2}-D)^{3}},\qquad D=Y_{ij}x_{i}x_{j},\qquad Y_{ij}=\frac{1}{2}\left[m_{i}^{2}+m_{j}^{2}-\Delta_{q_{i}q_{j}}\right]. \tag*{(2)} i M = ∫ ( 2 π ) 4 d 4 q ∭ d x 1 d x 2 d x 3 ( q 2 − D ) 3 2 δ ( ∑ i = 1 3 x i − 1 ) , D = Y ij x i x j , Y ij = 2 1 [ m i 2 + m j 2 − Δ q i q j ] . (2)
where
Δ q i q j = Δ q j q i , Δ q 1 q 2 = p 1 2 , Δ q 2 q 3 = p 3 2 , Δ q 3 q 1 = p 2 2 . (3) \Delta_{q_{i}q_{j}}=\Delta_{q_{j}q_{i}},\qquad\Delta_{q_{1}q_{2}}=p_{1}^{2},\qquad\Delta_{q_{2}q_{3}}=p_{3}^{2},\qquad\Delta_{q_{3}q_{1}}=p_{2}^{2}. \tag*{(3)} Δ q i q j = Δ q j q i , Δ q 1 q 2 = p 1 2 , Δ q 2 q 3 = p 3 2 , Δ q 3 q 1 = p 2 2 . (3)
The location of the triangle singularity is determined by the Landau equations [1, 11, 16]. Denoting s 1 = p 1 2 s_{1}=p_{1}^{2} s 1 = p 1 2 , s 2 = p 2 2 s_{2}=p_{2}^{2} s 2 = p 2 2 , and s 3 = p 3 2 s_{3}=p_{3}^{2} s 3 = p 3 2 , with s 1 s_{1} s 1 and s 2 s_{2} s 2 fixed, s 3 s_{3} s 3 satisfies the following equation:
det ( Y i j ) = 0 ⟹ 1 4 { m 1 2 [ m 2 2 ( − s 1 + s 2 + s 3 ) + m 3 2 ( s 1 − s 2 + s 3 ) + ( s 1 + s 2 − s 3 ) s 3 ] − s 3 m 1 4 − s 2 m 2 4 − s 1 [ m 2 2 ( s 1 − s 2 − s 3 ) + m 3 4 + s 2 s 3 ] + m 2 2 [ m 2 2 ( s 1 + s 2 − s 3 ) + s 2 ( s 1 − s 2 + s 3 ) ] } = 0. (4) \det(Y_{ij})=0\ \Longrightarrow\ \frac{1}{4}\left\{m_{1}^{2}\left[m_{2}^{2}(-s_{1}+s_{2}+s_{3})+m_{3}^{2}(s_{1}-s_{2}+s_{3})+(s_{1}+s_{2}-s_{3})s_{3}\right]-s_{3}m_{1}^{4}-s_{2}m_{2}^{4}-s_{1}\left[m_{2}^{2}(s_{1}-s_{2}-s_{3})+m_{3}^{4}+s_{2}s_{3}\right]+m_{2}^{2}\left[m_{2}^{2}(s_{1}+s_{2}-s_{3})+s_{2}(s_{1}-s_{2}+s_{3})\right]\right\}=0. \tag*{(4)} det ( Y ij ) = 0 ⟹ 4 1 { m 1 2 [ m 2 2 ( − s 1 + s 2 + s 3 ) + m 3 2 ( s 1 − s 2 + s 3 ) + ( s 1 + s 2 − s 3 ) s 3 ] − s 3 m 1 4 − s 2 m 2 4 − s 1 [ m 2 2 ( s 1 − s 2 − s 3 ) + m 3 4 + s 2 s 3 ] + m 2 2 [ m 2 2 ( s 1 + s 2 − s 3 ) + s 2 ( s 1 − s 2 + s 3 ) ] } = 0. (4)
The equation for s 3 s_{3} s 3 has two solutions. In the kinematic window considered here, the physical TS corresponds to the branch s 3 − s_{3}^{-} s 3 − that satisfies the Coleman–Norton condition, rather than merely lying within the kinematically allowed phase space:
s 3 = s 3 − = ( m 2 + m 3 ) 2 + 1 2 m 1 2 [ ( m 1 2 + m 3 2 − s 2 ) ( s 1 − m 1 2 − m 2 2 ) − 4 m 1 2 m 2 m 3 − λ [ s 1 , m 1 2 , m 2 2 ] λ [ s 2 , m 1 2 , m 3 2 ] ] . (5) s_{3}=s_{3}^{-}=(m_{2}+m_{3})^{2}+\frac{1}{2m_{1}^{2}}\left[(m_{1}^{2}+m_{3}^{2}-s_{2})(s_{1}-m_{1}^{2}-m_{2}^{2})-4m_{1}^{2}m_{2}m_{3}-\sqrt{\lambda[s_{1},m_{1}^{2},m_{2}^{2}]\lambda[s_{2},m_{1}^{2},m_{3}^{2}]}\right]. \tag*{(5)} s 3 = s 3 − = ( m 2 + m 3 ) 2 + 2 m 1 2 1 [ ( m 1 2 + m 3 2 − s 2 ) ( s 1 − m 1 2 − m 2 2 ) − 4 m 1 2 m 2 m 3 − λ [ s 1 , m 1 2 , m 2 2 ] λ [ s 2 , m 1 2 , m 3 2 ] ] . (5)
where λ ( x , y , z ) = x 2 + y 2 + z 2 − 2 x y − 2 y z − 2 z x \lambda(x,y,z)=x^2+y^2+z^2-2xy-2yz-2zx λ ( x , y , z ) = x 2 + y 2 + z 2 − 2 x y − 2 y z − 2 z x is the Källén triangle function [46 ] .
Although the position of the triangle singularity is fixed, different dynamical couplings can significantly affect the magnitude of the triangle singularity and even cancel its contribution. To investigate the impact of different dynamical couplings on the triangle singularity, we first consider the intermediate particles as ground-state S S S -wave particles with quantum numbers J P = 0 − , 1 − J^P=0^-,1^- J P = 0 − , 1 − . For the quantum numbers of the initial and final states, we consider S S S -wave or P P P -wave couplings of ground-state particles, leading to the following possibilities. The kinematic relations are
p 1 = q 1 + q 2 , q 1 = q 3 + p 2 , p 3 = q 2 + q 3 (6) p_1=q_1+q_2,\qquad q_1=q_3+p_2,\qquad p_3=q_2+q_3 \tag*{(6)} p 1 = q 1 + q 2 , q 1 = q 3 + p 2 , p 3 = q 2 + q 3 (6)
where p 1 , 2 , 3 p_{1,2,3} p 1 , 2 , 3 denote the initial and final state particles, q 1 , 2 , 3 q_{1,2,3} q 1 , 2 , 3 denote the intermediate particles, and V 1 , 2 , 3 V_{1,2,3} V 1 , 2 , 3 represent the partial wave types of the vertex couplings p 1 q 1 q 2 p_1q_1q_2 p 1 q 1 q 2 , p 2 q 1 q 3 p_2q_1q_3 p 2 q 1 q 3 , and p 3 q 2 q 3 p_3q_2q_3 p 3 q 2 q 3 , respectively. For the process p 3 → k 1 + k 2 p_3\to k_1+k_2 p 3 → k 1 + k 2 , which depends only on the external momenta, we focus on the cases 1 + → 1 − + 0 − 1^+\to1^-+0^- 1 + → 1 − + 0 − , 0 + → 0 − + 0 − 0^+\to0^-+0^- 0 + → 0 − + 0 − , 1 − → 0 − + 0 − 1^-\to0^-+0^- 1 − → 0 − + 0 − , and 0 − → 1 − + 0 − 0^-\to1^-+0^- 0 − → 1 − + 0 − .
Table 1. Possible quantum number combinations for the triangle diagram processes
We consider four types of particles: S S S (scalar), P P P (pseudoscalar), V V V (vector), and A A A (axial-vector). The amplitudes are labeled by their serial numbers. The coupling types involved are P P S PPS P P S , V P P VPP V P P , V V P VVP V V P , V P A VPA V P A , V V V VVV V V V , V V A VVA V V A , and V V S VVS V V S . The corresponding effective Lagrangians are given in Refs. [37 , 38 , 39 , 40 ] :
L P P S = g P P S P 1 P 2 S , L V P P = i g V P P ( P 1 ∂ μ P 2 − ∂ μ P 1 P 2 ) V μ , L V P A = g V P A V μ A μ P , \mathcal{L}_{PPS}=g_{PPS}P_1P_2S,\qquad\mathcal{L}_{VPP}=ig_{VPP}(P_1\partial_\mu P_2-\partial_\mu P_1P_2)V^\mu,\qquad\mathcal{L}_{VPA}=g_{VPA}V_\mu A^\mu P, L P P S = g P P S P 1 P 2 S , L V P P = i g V P P ( P 1 ∂ μ P 2 − ∂ μ P 1 P 2 ) V μ , L V P A = g V P A V μ A μ P ,
L V V P = g V V P ε α β μ ν ∂ α V 1 β ∂ μ V 2 ν P , L V V A = i g V V A ε α β μ ν ∂ α A β ∂ μ V 1 ν V 2 ν , \mathcal{L}_{VVP}=g_{VVP}\varepsilon_{\alpha\beta\mu\nu}\partial^\alpha V_1^\beta\partial^\mu V_2^\nu P,\qquad\mathcal{L}_{VVA}=ig_{VVA}\varepsilon_{\alpha\beta\mu\nu}\partial^\alpha A^\beta\partial^\mu V_1^\nu V_2^\nu, L V V P = g V V P ε α β μν ∂ α V 1 β ∂ μ V 2 ν P , L V V A = i g V V A ε α β μν ∂ α A β ∂ μ V 1 ν V 2 ν ,
L V V V = i g V V V ( V 1 , ν ∂ μ V 2 ν V 3 , μ + V 1 , μ V 2 ν ∂ μ V 3 , ν + V 2 , μ V 3 ν ∂ μ V 1 , ν ) , L V V S = g V V S V 1 μ V 2 μ S . (7) \mathcal{L}_{VVV}=ig_{VVV}(V_{1,\nu}\partial^\mu V_2^\nu V_{3,\mu}+V_{1,\mu}V_2^\nu\partial^\mu V_{3,\nu}+V_{2,\mu}V_3^\nu\partial^\mu V_{1,\nu}),\qquad\mathcal{L}_{VVS}=g_{VVS}V_1^\mu V_{2\mu}S. \tag*{(7)} L V V V = i g V V V ( V 1 , ν ∂ μ V 2 ν V 3 , μ + V 1 , μ V 2 ν ∂ μ V 3 , ν + V 2 , μ V 3 ν ∂ μ V 1 , ν ) , L V V S = g V V S V 1 μ V 2 μ S . (7)
The masses of the intermediate particles in the loop diagram are denoted by m 1 , 2 , 3 m_{1,2,3} m 1 , 2 , 3 , and the masses of the external particles p 1 , 2 , 3 p_{1,2,3} p 1 , 2 , 3 by M 1 , 2 , 3 M_{1,2,3} M 1 , 2 , 3 . The invariant mass squared of the final state is s 3 = p 3 2 = ( k 1 + k 2 ) 2 s_3=p_3^2=(k_1+k_2)^2 s 3 = p 3 2 = ( k 1 + k 2 ) 2 . The amplitudes can be written as
i M 1 = g 1 ∫ d 4 q 3 ( 2 π ) 4 ε 1 i μ g μ ν ( q 3 ν − p 2 ν ) ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) 1 s 3 − M 3 2 + i M 3 Γ 3 , (8) i\mathcal{M}_1=g_1\int\frac{d^4q_3}{(2\pi)^4}\frac{\varepsilon^\mu_{1i}g_{\mu\nu}(q_3^\nu-p_2^\nu)}{(q_1^2-m_1^2)(q_2^2-m_2^2)(q_3^2-m_3^2)}\frac{1}{s_3-M_3^2+iM_3\Gamma_3}, \tag*{(8)} i M 1 = g 1 ∫ ( 2 π ) 4 d 4 q 3 ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) ε 1 i μ g μν ( q 3 ν − p 2 ν ) s 3 − M 3 2 + i M 3 Γ 3 1 , (8)
i M 2 = g 2 ∫ d 4 q 3 ( 2 π ) 4 ε 1 i μ g μ ν ( p 2 − q 3 ) ν ( q 2 − q 3 ) α g α β ( k 1 − k 2 ) β ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) 1 s 3 − M 3 2 + i M 3 Γ 3 , (9) i\mathcal{M}_2=g_2\int\frac{d^4q_3}{(2\pi)^4}\frac{\varepsilon^\mu_{1i}g_{\mu\nu}(p_2-q_3)^\nu(q_2-q_3)^\alpha g_{\alpha\beta}(k_1-k_2)^\beta}{(q_1^2-m_1^2)(q_2^2-m_2^2)(q_3^2-m_3^2)}\frac{1}{s_3-M_3^2+iM_3\Gamma_3}, \tag*{(9)} i M 2 = g 2 ∫ ( 2 π ) 4 d 4 q 3 ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) ε 1 i μ g μν ( p 2 − q 3 ) ν ( q 2 − q 3 ) α g α β ( k 1 − k 2 ) β s 3 − M 3 2 + i M 3 Γ 3 1 , (9)
i M 3 = g 3 ∫ d 4 q 3 ( 2 π ) 4 ε μ ν ρ σ ε 1 i μ p 1 ν ( p 2 − q 3 ) ρ g σ δ ε λ τ δ η q 2 τ p 3 η g α β ( k 1 − k 2 ) β ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) 1 s 3 − M 3 2 + i M 3 Γ 3 , (10) i\mathcal{M}_3=g_3\int\frac{d^4q_3}{(2\pi)^4}\frac{\varepsilon_{\mu\nu\rho\sigma}\varepsilon^\mu_{1i}p_1^\nu(p_2-q_3)^\rho g^\sigma{}_\delta\varepsilon_{\lambda\tau\delta\eta}q_2^\tau p_3^\eta g^{\alpha\beta}(k_1-k_2)_\beta}{(q_1^2-m_1^2)(q_2^2-m_2^2)(q_3^2-m_3^2)}\frac{1}{s_3-M_3^2+iM_3\Gamma_3}, \tag*{(10)} i M 3 = g 3 ∫ ( 2 π ) 4 d 4 q 3 ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) ε μν ρ σ ε 1 i μ p 1 ν ( p 2 − q 3 ) ρ g σ δ ε λ τ δ η q 2 τ p 3 η g α β ( k 1 − k 2 ) β s 3 − M 3 2 + i M 3 Γ 3 1 , (10)
i M 4 = g 4 ∫ d 4 q 3 ( 2 π ) 4 ε μ ν ρ σ ε 1 i μ p 1 ν ( p 2 − q 3 ) ρ ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) ε 1 f σ s 3 − M 3 2 + i M 3 Γ 3 , (11) i\mathcal{M}_4=g_4\int\frac{d^4q_3}{(2\pi)^4}\frac{\varepsilon_{\mu\nu\rho\sigma}\varepsilon^\mu_{1i}p_1^\nu(p_2-q_3)^\rho}{(q_1^2-m_1^2)(q_2^2-m_2^2)(q_3^2-m_3^2)}\frac{\varepsilon^\sigma_{1f}}{s_3-M_3^2+iM_3\Gamma_3}, \tag*{(11)} i M 4 = g 4 ∫ ( 2 π ) 4 d 4 q 3 ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) ε μν ρ σ ε 1 i μ p 1 ν ( p 2 − q 3 ) ρ s 3 − M 3 2 + i M 3 Γ 3 ε 1 f σ , (11)
i M 5 = g 5 ∫ d 4 q 3 ( 2 π ) 4 ε 1 i α ( g ρ σ ( q 2 − q 1 ) μ − g ρ σ ( p 1 + q 2 ) ν + g μ ρ g σ α ( q 1 + p 1 ) α ) ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) ( p 2 − q 3 ) ρ ( p 3 + q 3 ) σ ε 1 f β ( k 2 + p 3 ) β s 3 − M 3 2 + i M 3 Γ 3 . (12) i\mathcal{M}_5=g_5\int\frac{d^4q_3}{(2\pi)^4}\frac{\varepsilon^\alpha_{1i}\left(g_{\rho\sigma}(q_2-q_1)_\mu-g_{\rho\sigma}(p_1+q_2)^\nu+g_{\mu\rho}g_{\sigma\alpha}(q_1+p_1)^\alpha\right)}{(q_1^2-m_1^2)(q_2^2-m_2^2)(q_3^2-m_3^2)}
\frac{(p_2-q_3)^\rho(p_3+q_3)^\sigma\varepsilon^\beta_{1f}(k_2+p_3)_\beta}{s_3-M_3^2+iM_3\Gamma_3}. \tag*{(12)} i M 5 = g 5 ∫ ( 2 π ) 4 d 4 q 3 ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) ε 1 i α ( g ρ σ ( q 2 − q 1 ) μ − g ρ σ ( p 1 + q 2 ) ν + g μ ρ g σ α ( q 1 + p 1 ) α ) s 3 − M 3 2 + i M 3 Γ 3 ( p 2 − q 3 ) ρ ( p 3 + q 3 ) σ ε 1 f β ( k 2 + p 3 ) β . (12)
i M 6 = g 6 ∫ d 4 q 3 ( 2 π ) 4 ε μ ν ρ σ ε i 1 μ p 1 ν g 1 ρ g 1 σ ( p 2 − q 3 ) λ ( q 2 − q 3 ) τ ( k 1 − k 2 ) τ ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) ( s 3 − M 3 2 + i M 3 Γ 3 ) (13) i\mathcal{M}_6=g_6\int\frac{d^4q_3}{(2\pi)^4}\frac{\varepsilon_{\mu\nu\rho\sigma}\varepsilon_{i1}^{\mu}p_1^\nu g_1^\rho g_1^\sigma(p_2-q_3)_\lambda(q_2-q_3)^\tau(k_1-k_2)_\tau}{(q_1^2-m_1^2)(q_2^2-m_2^2)(q_3^2-m_3^2)(s_3-M_3^2+iM_3\Gamma_3)} \tag*{(13)} i M 6 = g 6 ∫ ( 2 π ) 4 d 4 q 3 ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) ( s 3 − M 3 2 + i M 3 Γ 3 ) ε μν ρ σ ε i 1 μ p 1 ν g 1 ρ g 1 σ ( p 2 − q 3 ) λ ( q 2 − q 3 ) τ ( k 1 − k 2 ) τ (13)
i M 7 = g 7 ∫ d 4 q 3 ( 2 π ) 4 ε i 1 μ ( g ρ σ ( q 2 − q 1 ) μ − g ν ρ g μ σ ( p 1 + q 2 ) ν + g μ ρ g α σ ( q 1 + p 1 ) α ) ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) ( p 2 − q 3 ) ρ g σ η ε η δ λ τ q 2 δ p 3 λ ( k 1 − k 2 ) τ s 3 − M 3 2 + i M 3 Γ 3 (14) i\mathcal{M}_7=g_7\int\frac{d^4q_3}{(2\pi)^4}\frac{\varepsilon_{i1}^{\mu}\left(g_{\rho\sigma}(q_2-q_1)_\mu-g_{\nu\rho}g_{\mu\sigma}(p_1+q_2)^\nu+g_{\mu\rho}g_{\alpha\sigma}(q_1+p_1)^\alpha\right)}{(q_1^2-m_1^2)(q_2^2-m_2^2)(q_3^2-m_3^2)}
\frac{(p_2-q_3)^\rho g^{\sigma\eta}\varepsilon_{\eta\delta\lambda\tau}q_2^\delta p_3^\lambda(k_1-k_2)^\tau}{s_3-M_3^2+iM_3\Gamma_3} \tag*{(14)} i M 7 = g 7 ∫ ( 2 π ) 4 d 4 q 3 ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) ε i 1 μ ( g ρ σ ( q 2 − q 1 ) μ − g ν ρ g μ σ ( p 1 + q 2 ) ν + g μ ρ g α σ ( q 1 + p 1 ) α ) s 3 − M 3 2 + i M 3 Γ 3 ( p 2 − q 3 ) ρ g σ η ε η δ λ τ q 2 δ p 3 λ ( k 1 − k 2 ) τ (14)
i M 8 = g 8 ∫ d 4 q 3 ( 2 π ) 4 ε i 1 μ ( g ρ σ ( q 2 − q 1 ) μ − g ν ρ g μ σ ( p 1 + q 2 ) ν + g μ ρ g α σ ( q 1 + p 1 ) α ) ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) ( p 2 − q 3 ) ρ ε 1 f ∗ σ s 3 − M 3 2 + i M 3 Γ 3 (15) i\mathcal{M}_8=g_8\int\frac{d^4q_3}{(2\pi)^4}\frac{\varepsilon_{i1}^{\mu}\left(g_{\rho\sigma}(q_2-q_1)_\mu-g_{\nu\rho}g_{\mu\sigma}(p_1+q_2)^\nu+g_{\mu\rho}g_{\alpha\sigma}(q_1+p_1)^\alpha\right)}{(q_1^2-m_1^2)(q_2^2-m_2^2)(q_3^2-m_3^2)}
\frac{(p_2-q_3)^\rho\varepsilon_{1f}^{*\sigma}}{s_3-M_3^2+iM_3\Gamma_3} \tag*{(15)} i M 8 = g 8 ∫ ( 2 π ) 4 d 4 q 3 ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) ε i 1 μ ( g ρ σ ( q 2 − q 1 ) μ − g ν ρ g μ σ ( p 1 + q 2 ) ν + g μ ρ g α σ ( q 1 + p 1 ) α ) s 3 − M 3 2 + i M 3 Γ 3 ( p 2 − q 3 ) ρ ε 1 f ∗ σ (15)
i M 9 = g 9 ∫ d 4 q 3 ( 2 π ) 4 ( p 2 − q 3 ) μ ε 1 f μ ∗ ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) 1 s 3 − M 3 2 + i M 3 Γ 3 (16) i\mathcal{M}_9=g_9\int\frac{d^4q_3}{(2\pi)^4}\frac{(p_2-q_3)^\mu\varepsilon_{1f\mu}^*}{(q_1^2-m_1^2)(q_2^2-m_2^2)(q_3^2-m_3^2)}\frac{1}{s_3-M_3^2+iM_3\Gamma_3} \tag*{(16)} i M 9 = g 9 ∫ ( 2 π ) 4 d 4 q 3 ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) ( p 2 − q 3 ) μ ε 1 f μ ∗ s 3 − M 3 2 + i M 3 Γ 3 1 (16)
i M 10 = g 10 ∫ d 4 q 3 ( 2 π ) 4 ( p 2 − q 3 ) μ ( p 1 + q 2 ) μ ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) 1 s 3 − M 3 2 + i M 3 Γ 3 (17) i\mathcal{M}_{10}=g_{10}\int\frac{d^4q_3}{(2\pi)^4}\frac{(p_2-q_3)^\mu(p_1+q_2)_\mu}{(q_1^2-m_1^2)(q_2^2-m_2^2)(q_3^2-m_3^2)}\frac{1}{s_3-M_3^2+iM_3\Gamma_3} \tag*{(17)} i M 10 = g 10 ∫ ( 2 π ) 4 d 4 q 3 ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) ( p 2 − q 3 ) μ ( p 1 + q 2 ) μ s 3 − M 3 2 + i M 3 Γ 3 1 (17)
i M 11 = g 11 ∫ d 4 q 3 ( 2 π ) 4 ( p 1 + q 2 ) μ ( p 2 − q 3 ) μ ( q 3 − q 2 ) ν ( k 1 − k 2 ) ν ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) 1 s 3 − M 3 2 + i M 3 Γ 3 (18) i\mathcal{M}_{11}=g_{11}\int\frac{d^4q_3}{(2\pi)^4}\frac{(p_1+q_2)^\mu(p_2-q_3)_\mu(q_3-q_2)^\nu(k_1-k_2)_\nu}{(q_1^2-m_1^2)(q_2^2-m_2^2)(q_3^2-m_3^2)}\frac{1}{s_3-M_3^2+iM_3\Gamma_3} \tag*{(18)} i M 11 = g 11 ∫ ( 2 π ) 4 d 4 q 3 ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) ( p 1 + q 2 ) μ ( p 2 − q 3 ) μ ( q 3 − q 2 ) ν ( k 1 − k 2 ) ν s 3 − M 3 2 + i M 3 Γ 3 1 (18)
i M 12 = g 12 ∫ d 4 q 3 ( 2 π ) 4 ε μ ν ρ σ q 1 μ q 2 ν ( p 2 − q 3 ) ρ g α σ ε α λ δ τ q 2 λ p 3 τ ( k 1 − k 2 ) δ ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) 1 s 3 − M 3 2 + i M 3 Γ 3 (19) i\mathcal{M}_{12}=g_{12}\int\frac{d^4q_3}{(2\pi)^4}\frac{\varepsilon_{\mu\nu\rho\sigma}q_1^\mu q_2^\nu(p_2-q_3)^\rho g^{\alpha\sigma}\varepsilon_{\alpha\lambda\delta\tau}q_2^\lambda p_3^\tau(k_1-k_2)^\delta}{(q_1^2-m_1^2)(q_2^2-m_2^2)(q_3^2-m_3^2)}\frac{1}{s_3-M_3^2+iM_3\Gamma_3} \tag*{(19)} i M 12 = g 12 ∫ ( 2 π ) 4 d 4 q 3 ( q 1 2 − m 1 2 ) ( q 2 2 − m 2 2 ) ( q 3 2 − m 3 2 ) ε μν ρ σ q 1 μ q 2 ν ( p 2 − q 3 ) ρ g α σ ε α λ δ τ q 2 λ p 3 τ ( k 1 − k 2 ) δ s 3 − M 3 2 + i M 3 Γ 3 1 (19)
Here, g i g_i g i (i = 1 , 2 , ⋯ , 12 i=1,2,\cdots,12 i = 1 , 2 , ⋯ , 12 ) denotes the product of coupling constants involved in the loop diagram. ε i 1 \varepsilon_{i1} ε i 1 is the polarization vector of the initial-state particle, and ε 1 f ∗ \varepsilon_{1f}^* ε 1 f ∗ is the polarization vector of the final-state particle.
The one-loop amplitudes can be decomposed into scalar loop integrals with Lorentz-scalar coefficients [45 ] [16].
i M i g i = ( ∑ j A i j A 0 j + ∑ j B i j B 0 j + C i C 0 ) 1 s 3 − M 3 2 + i M 3 Γ 3 (20) \frac{i\mathcal{M}_i}{g_i}=\left(\sum_j A_i^j A_{0j}+\sum_j B_i^j B_{0j}+C_iC_0\right)\frac{1}{s_3-M_3^2+iM_3\Gamma_3} \tag*{(20)} g i i M i = ( j ∑ A i j A 0 j + j ∑ B i j B 0 j + C i C 0 ) s 3 − M 3 2 + i M 3 Γ 3 1 (20)
The scalar loop integrals A 0 j \mathbf{A}_{0j} A 0 j , B 0 j \mathbf{B}_{0j} B 0 j , and C 0 \mathbf{C}_0 C 0 are defined as
A 01 = ∫ d 4 q 3 ( 2 π ) 4 1 D 1 , A 02 = ∫ d 4 q 3 ( 2 π ) 4 1 D 2 , A 03 = ∫ d 4 q 3 ( 2 π ) 4 1 D 3 , A_{01}=\int\frac{d^4q_3}{(2\pi)^4}\frac{1}{D_1},\qquad A_{02}=\int\frac{d^4q_3}{(2\pi)^4}\frac{1}{D_2},\qquad A_{03}=\int\frac{d^4q_3}{(2\pi)^4}\frac{1}{D_3}, A 01 = ∫ ( 2 π ) 4 d 4 q 3 D 1 1 , A 02 = ∫ ( 2 π ) 4 d 4 q 3 D 2 1 , A 03 = ∫ ( 2 π ) 4 d 4 q 3 D 3 1 ,
B 01 = ∫ d 4 q 3 ( 2 π ) 4 1 D 2 D 3 , B 02 = ∫ d 4 q 3 ( 2 π ) 4 1 D 1 D 3 , B 03 = ∫ d 4 q 3 ( 2 π ) 4 1 D 1 D 2 , B_{01}=\int\frac{d^4q_3}{(2\pi)^4}\frac{1}{D_2D_3},\qquad B_{02}=\int\frac{d^4q_3}{(2\pi)^4}\frac{1}{D_1D_3},\qquad B_{03}=\int\frac{d^4q_3}{(2\pi)^4}\frac{1}{D_1D_2}, B 01 = ∫ ( 2 π ) 4 d 4 q 3 D 2 D 3 1 , B 02 = ∫ ( 2 π ) 4 d 4 q 3 D 1 D 3 1 , B 03 = ∫ ( 2 π ) 4 d 4 q 3 D 1 D 2 1 ,
C 0 = ∫ d 4 q 3 ( 2 π ) 4 1 D 1 D 2 D 3 (21) C_0=\int\frac{d^4q_3}{(2\pi)^4}\frac{1}{D_1D_2D_3} \tag*{(21)} C 0 = ∫ ( 2 π ) 4 d 4 q 3 D 1 D 2 D 3 1 (21)
where D 1 = q 1 2 − m 1 2 D_1=q_1^2-m_1^2 D 1 = q 1 2 − m 1 2 , D 2 = q 2 2 − m 2 2 D_2=q_2^2-m_2^2 D 2 = q 2 2 − m 2 2 , and D 3 = q 3 2 − m 3 2 D_3=q_3^2-m_3^2 D 3 = q 3 2 − m 3 2 . The explicit expressions for A i j A_i^j A i j , B i j B_i^j B i j , and C i C_i C i can be found in Appendix § 1.2.1 (see Tables V to Table XVI ). We also consider the case without an intermediate resonance in the final state, i.e., q 2 q 3 → k 1 k 2 q_2q_3\to k_1k_2 q 2 q 3 → k 1 k 2 proceeds via a contact interaction. In this case, the corresponding amplitude is obtained by removing the Breit-Wigner factor, and the amplitude is given by
Table XVI. TABLE XVI: Coefficients of scalar loop integrals in the decomposition of i M 12 i\mathcal{M}_{12} i M 12
i M i C = i M i × ( s 3 − M 3 2 + i M 3 Γ 3 ) . (22) i\mathcal{M}_i^{\mathrm{C}}=i\mathcal{M}_i\times\left(s_3-M_3^2+iM_3\Gamma_3\right). \tag*{(22)} i M i C = i M i × ( s 3 − M 3 2 + i M 3 Γ 3 ) . (22)
To regulate the ultraviolet behavior of the loop integrals, we introduce the form factor
F ( q 2 ) = ∏ i = 1 3 Λ i 2 − m i 2 Λ i 2 − q i 2 , (23) \mathcal{F}(q^2)=\prod_{i=1}^{3}\frac{\Lambda_i^2-m_i^2}{\Lambda_i^2-q_i^2}, \tag*{(23)} F ( q 2 ) = i = 1 ∏ 3 Λ i 2 − q i 2 Λ i 2 − m i 2 , (23)
where Λ i = m i + α Λ Q C D \Lambda_i=m_i+\alpha\Lambda_{\mathrm{QCD}} Λ i = m i + α Λ QCD , with Λ Q C D = 0.22 \Lambda_{\mathrm{QCD}}=0.22 Λ QCD = 0.22 GeV denoting the QCD energy scale and α \alpha α the cutoff parameter. For small values of the cutoff parameter, the form factor may introduce additional singularities. We therefore choose α = 2 \alpha=2 α = 2 , so that the singularities induced by the form factor remain far from the energy region of interest. For brevity, the form factor F ( q 2 ) \mathcal{F}(q^2) F ( q 2 ) is not written explicitly in the amplitude expressions above, but it is included in the complete amplitudes used to obtain the spectra and the relative contributions in Table III . The coefficient scans and the separate scalar-integral plots shown below instead use the physical masses without the form-factor subtraction, as specified in the corresponding captions. The form factor can be implemented by mass subtraction, since
Table III. Numerical results for a i j k a_i^{jk} a i j k , b i j k b_i^{jk} b i j k , and ∣ R [ f C j k C 0 ] ∣ \left|\mathcal{R}[f_{C_j}^k C_0]\right| R [ f C j k C 0 ] with the form factor included. Parentheses list the independent Lorentz components in the order of Table II.
1 q r 2 − m r 2 Λ r 2 − m r 2 Λ r 2 − q r 2 = 1 q r 2 − m r 2 − 1 q r 2 − Λ r 2 . (24) \frac{1}{q_r^2-m_r^2}\frac{\Lambda_r^2-m_r^2}{\Lambda_r^2-q_r^2}=\frac{1}{q_r^2-m_r^2}-\frac{1}{q_r^2-\Lambda_r^2}. \tag*{(24)} q r 2 − m r 2 1 Λ r 2 − q r 2 Λ r 2 − m r 2 = q r 2 − m r 2 1 − q r 2 − Λ r 2 1 . (24)
We therefore define the eight-term subtraction operator
R [ X ] = ( 1 − T 1 ) ( 1 − T 2 ) ( 1 − T 3 ) X , T r X = X ∣ m r 2 → Λ r 2 . (25) \mathcal{R}[X]=(1-T_1)(1-T_2)(1-T_3)X,\qquad T_rX=\left.X\right|_{m_r^2\to\Lambda_r^2}. \tag*{(25)} R [ X ] = ( 1 − T 1 ) ( 1 − T 2 ) ( 1 − T 3 ) X , T r X = X ∣ m r 2 → Λ r 2 . (25)
Here each replacement acts on the corresponding propagator-mass parameter, with the external kinematics, couplings, and cutoff values held fixed. After the scalar-integral decomposition, R \mathcal{R} R must act on each complete coefficient-times-integral product, rather than on the scalar integral alone. The three propagator masses are kept independent during the subtraction; physical mass equalities are imposed only afterwards. The analytic coefficients in Appendix A \mathrm{A} A are those of the unregulated amplitudes.
Numerical Results and Discussions
Spectra and scalar-integral decomposition
To compare the dynamical effects of different vertex structures, we use the masses and kinematics of the η ( 1405 ) → π + π − π 0 \eta(1405)\to\pi^+\pi^-\pi^0 η ( 1405 ) → π + π − π 0 system with a K ∗ K ˉ K K^*\bar{K}K K ∗ K ˉ K loop as a common numerical benchmark for the twelve cases in Table I. Only case 10 has the spin-parity assignments of that physical decay chain with an intermediate f 0 ( 980 ) f_0(980) f 0 ( 980 ) ; the other cases retain the same mass inputs while varying the spins and vertex couplings. The f 0 ( 980 ) f_0(980) f 0 ( 980 ) mass and width are used as common parameters in the final-state resonance factor, without identifying the intermediate resonance in every case with the physical f 0 ( 980 ) f_0(980) f 0 ( 980 ) . The resulting k 1 k 2 k_1k_2 k 1 k 2 invariant-mass spectra, with and without this resonance factor, are shown in Figure 2 and Figure 3 .
Figure 2. Invariant-mass spectra s 3 = M ( k 1 k 2 ) \sqrt{s_3}=M(k_1k_2) s 3 = M ( k 1 k 2 ) for the twelve spin-parity assignments in Table I, using the common mass inputs and including the final-state resonance factor with the f 0 ( 980 ) f_0(980) f 0 ( 980 ) parameters.
Figure 3. Invariant-mass spectra s 3 = M ( k 1 k 2 ) \sqrt{s_3}=M(k_1k_2) s 3 = M ( k 1 k 2 ) for the same twelve spin-parity assignments and mass inputs as in Figure 2 , with the final-state resonance factor removed.
From these invariant-mass spectra, we see that the prominence of the triangle singularity varies with the choice of vertex couplings. In general, a dominant C 0 C_0 C 0 contribution with a nonvanishing Landau-logarithmic coefficient produces a narrow peak. Dynamical effects can suppress this coefficient and weaken the triangle enhancement. The resulting cusp-like profiles must be understood through the remaining ordinary-threshold dependence, smooth background, and interference in the complete amplitude, rather than by assuming that an explicit B 0 B_0 B 0 term becomes dominant. We therefore next analyze the scalar-integral contributions together with their coherent threshold expansion to assess the dynamical effects on the triangle singularity.
After the scalar-integral reduction, the coefficients A i j A_i^j A i j , B i j B_i^j B i j , and C i C_i C i are linear combinations of Lorentz scalars formed from the external momenta and polarization vectors. The coefficients of these Lorentz structures depend on the internal masses and external kinematic invariants:
A i j = ∑ k L i k f A i j k , B i j = ∑ k L i k f B i j k , C i = ∑ k L i k f C i k , (26) A_i^j=\sum_k L_i^k f_{Ai}^{jk},\qquad B_i^j=\sum_k L_i^k f_{Bi}^{jk},\qquad C_i=\sum_k L_i^k f_{Ci}^k, \tag*{(26)} A i j = k ∑ L i k f A i j k , B i j = k ∑ L i k f B i j k , C i = k ∑ L i k f C i k , (26)
where L i k L_i^k L i k is the k k k -th independent Lorentz scalar in the i i i -th diagram, and f A i j k f_{Ai}^{jk} f A i j k , f B i j k f_{Bi}^{jk} f B i j k , and f C i k f_{Ci}^k f C i k are its coefficients multiplying the corresponding scalar integrals. By comparing the relative sizes of ∣ R [ f A i j k A 0 j ] ∣ |\mathcal{R}[f_{Ai}^{jk}A_{0j}]| ∣ R [ f A i j k A 0 j ] ∣ and ∣ R [ f B i j k B 0 j ] ∣ |\mathcal{R}[f_{Bi}^{jk}B_{0j}]| ∣ R [ f B i j k B 0 j ] ∣ with ∣ R [ f C i k C 0 ] ∣ |\mathcal{R}[f_{Ci}^kC_{0}]| ∣ R [ f C i k C 0 ] ∣ , we can determine the regulated contributions of the different scalar functions and thereby analyze the dynamical effects on the triangle singularity. We first list all the possible Lorentz scalars obtained from the decomposition of the diagrams, as shown in Table II . To compare the contributions of different scalar functions near the triangle singularity, we apply the form-factor subtraction to each complete coefficient-times-integral product and define
Table II. The possible Lorentz scalars obtained from the decomposition of the diagrams
a i j k = ∣ R [ f A i j k A 0 j ] ∣ ∣ R [ f C i k C 0 ] ∣ , b i j k = ∣ R [ f B i j k B 0 j ] ∣ ∣ R [ f C i k C 0 ] ∣ . (27) a_i^{jk}=\frac{\left|\mathcal{R}[f_{Ai}^{jk}A_{0j}]\right|}{\left|\mathcal{R}[f_{Ci}^{k}C_{0}]\right|},\qquad b_i^{jk}=\frac{\left|\mathcal{R}[f_{Bi}^{jk}B_{0j}]\right|}{\left|\mathcal{R}[f_{Ci}^{k}C_{0}]\right|}. \tag*{(27)} a i j k = R [ f C i k C 0 ] R [ f A i j k A 0 j ] , b i j k = R [ f C i k C 0 ] R [ f B i j k B 0 j ] . (27)
The numerator and denominator are regulated separately before taking their absolute values. Under the adopted three-propagator mass-subtraction prescription, each A 0 A_0 A 0 or B 0 B_0 B 0 integral depends on only one or two internal masses, respectively. If the coefficient multiplying such an integral is independent of the internal masses, the complete product is independent of at least one subtracted mass. The corresponding factor ( 1 − T r ) (1-T_r) ( 1 − T r ) in Eq. (25) therefore cancels this product exactly. For the amplitudes considered here, all A 0 A_0 A 0 coefficients satisfy this condition, giving
R [ f A i j k A 0 j ] = 0. (28) \mathcal{R}[f_{Ai}^{jk}A_{0j}]=0. \tag*{(28)} R [ f A i j k A 0 j ] = 0. (28)
This cancellation explains the vanishing a i j k a_i^{jk} a i j k and some of the exact zeros of b i j k b_i^{jk} b i j k in Table III , provided that the common normalization denominator is nonzero. Internal-mass dependence in a coefficient can prevent this cancellation, which is why the subtraction must act on the complete coefficient-times-integral product rather than on the bare scalar integral alone.
Threshold and Landau-point expansion
The normal two-particle threshold associated with the q 2 q 3 q_2q_3 q 2 q 3 intermediate state is
s t h = ( m 2 + m 3 ) 2 , (29) s_{\mathrm{th}}=(m_2+m_3)^2, \tag*{(29)} s th = ( m 2 + m 3 ) 2 , (29)
whereas the physical triangle singularity is located at s 3 = s ˉ 3 s_3=\bar{s}_3 s 3 = s ˉ 3 . These two points are close in the kinematic region considered here, but they are not identical in general. We therefore introduce a local variable for each expansion point,
δ = s 3 − s t h , Δ = s 3 − s ˉ 3 , G ( s 3 ) = λ ( s 1 , s 2 , s 3 ) . (30) \delta=s_3-s_{\mathrm{th}},\qquad\Delta=s_3-\bar{s}_3,\qquad G(s_3)=\lambda(s_1,s_2,s_3). \tag*{(30)} δ = s 3 − s th , Δ = s 3 − s ˉ 3 , G ( s 3 ) = λ ( s 1 , s 2 , s 3 ) . (30)
The relative momentum of the q 2 q 3 q_2q_3 q 2 q 3 pair satisfies
∣ q ∣ 2 = λ ( s 3 , m 2 2 , m 3 2 ) 4 s 3 = m 2 m 3 ( m 2 + m 3 ) 2 δ + O ( δ 2 ) , (31) |\mathbf{q}|^2=\frac{\lambda(s_3,m_2^2,m_3^2)}{4s_3}=\frac{m_2m_3}{(m_2+m_3)^2}\delta+O(\delta^2), \tag*{(31)} ∣ q ∣ 2 = 4 s 3 λ ( s 3 , m 2 2 , m 3 2 ) = ( m 2 + m 3 ) 2 m 2 m 3 δ + O ( δ 2 ) , (31)
so that this expansion also makes the threshold powers to be associated with the vertex structure explicitly.
At fixed s 1 s_1 s 1 and s 2 s_2 s 2 , the normal threshold and the Landau point require separate local expansions [9, 12]; exact scalar-integral expressions and their analytic continuations are given in Refs. [47 , 48 , 49 ] . Near an isolated normal threshold, both B 01 B_{01} B 01 and C 0 C_0 C 0 carry the q 2 q 3 q_2q_3 q 2 q 3 two-particle cut. At threshold, the on-shell momentum is q 3 = m 3 p 3 / ( m 2 + m 3 ) q_3=m_3p_3/(m_2+m_3) q 3 = m 3 p 3 / ( m 2 + m 3 ) , so the uncut propagator denominator becomes
d ∗ ≡ D 1 ∣ q 2 q 3 t h r e s h o l d = s 2 + m 3 2 − m 1 2 + m 3 m 2 + m 3 ( s 1 − s 2 − s t h ) . (32) d_{*} \equiv D_{1}|_{q_{2}q_{3}\ \mathrm{threshold}}=s_{2}+m_{3}^{2}-m_{1}^{2}+\frac{m_{3}}{m_{2}+m_{3}}(s_{1}-s_{2}-s_{\mathrm{th}}). \tag*{(32)} d ∗ ≡ D 1 ∣ q 2 q 3 threshold = s 2 + m 3 2 − m 1 2 + m 2 + m 3 m 3 ( s 1 − s 2 − s th ) . (32)
For d ∗ ≠ 0 d_{*}\ne0 d ∗ = 0 , applying the two-particle Cutkosky rule [3] shows that the leading triangle discontinuity equals the leading bubble discontinuity divided by d ∗ d_{*} d ∗ . Expanding the standard scalar-integral representations [48 ] about this isolated normal threshold, we obtain
B 01 ( s 3 ) = B 01 r e g , t h ( s 3 ) + ρ B − δ − i 0 + O ( ∣ δ ∣ 3 / 2 ) , C 0 ( s 3 ) = C 0 r e g , t h ( s 3 ) + ρ B d ∗ − δ − i 0 + O ( ∣ δ ∣ 3 / 2 ) . (33) \begin{aligned}
B_{01}(s_{3})&=B_{01}^{\mathrm{reg,th}}(s_{3})+\rho_{B}\sqrt{-\delta-i0}+O(|\delta|^{3/2}),\\
C_{0}(s_{3})&=C_{0}^{\mathrm{reg,th}}(s_{3})+\frac{\rho_{B}}{d_{*}}\sqrt{-\delta-i0}+O(|\delta|^{3/2}). \tag*{(33)}
\end{aligned} B 01 ( s 3 ) C 0 ( s 3 ) = B 01 reg , th ( s 3 ) + ρ B − δ − i 0 + O ( ∣ δ ∣ 3/2 ) , = C 0 reg , th ( s 3 ) + d ∗ ρ B − δ − i 0 + O ( ∣ δ ∣ 3/2 ) . (33)
The regular parts include the analytic terms at s t h s_{\mathrm{th}} s th , and the normalization-dependent coefficient ρ B \rho_{B} ρ B includes the phase appropriate to the chosen square-root branch. The neighborhood of this expansion excludes the distinct Landau point and other simultaneous thresholds. In the following one-variable notation, the dependence on fixed masses, s 1 s_{1} s 1 , and s 2 s_{2} s 2 is suppressed. The remaining independent invariant s 23 s_{23} s 23 is held fixed unless a smooth kinematic path is specified, and the Lorentz scalars are evaluated in a polarization basis chosen smoothly along the path. For each independent Lorentz scalar, define
C ^ i k ( s 3 ) = L i k ( s 3 ) f C i k ( s 3 ) , B ^ i k ( s 3 ) = L i k ( s 3 ) f B i k ( s 3 ) . (34) \widehat{C}_{i}^{k}(s_{3})=L_{i}^{k}(s_{3})f_{C_{i}}^{k}(s_{3}),\qquad\widehat{B}_{i}^{k}(s_{3})=L_{i}^{k}(s_{3})f_{B_{i}}^{k}(s_{3}). \tag*{(34)} C i k ( s 3 ) = L i k ( s 3 ) f C i k ( s 3 ) , B i k ( s 3 ) = L i k ( s 3 ) f B i k ( s 3 ) . (34)
Assuming that these coefficients and the external resonance factor are regular at s t h s_{\mathrm{th}} s th , the leading normal-threshold term of the unregulated amplitude is
i M i , k t h , n o n a n ( s 3 ) = g i ρ B s t h − M 3 2 + i M 3 Γ 3 [ B ^ i k ( s t h ) + C ^ i k ( s t h ) d ∗ ] − δ − i 0 + O ( ∣ δ ∣ 3 / 2 ) . (35) i\mathcal{M}_{i,k}^{\mathrm{th,nonan}}(s_{3})=\frac{g_{i}\rho_{B}}{s_{\mathrm{th}}-M_{3}^{2}+iM_{3}\Gamma_{3}}\left[\widehat{B}_{i}^{k}(s_{\mathrm{th}})+\frac{\widehat{C}_{i}^{k}(s_{\mathrm{th}})}{d_{*}}\right]\sqrt{-\delta-i0}+O(|\delta|^{3/2}). \tag*{(35)} i M i , k th , nonan ( s 3 ) = s th − M 3 2 + i M 3 Γ 3 g i ρ B [ B i k ( s th ) + d ∗ C i k ( s th ) ] − δ − i 0 + O ( ∣ δ ∣ 3/2 ) . (35)
Thus, its square-root coefficient is a coherent sum, not the coefficient of B 01 B_{01} B 01 alone. In particular, with m 3 = m 2 m_{3}=m_{2} m 3 = m 2 , one has s t h = 4 m 2 2 s_{\mathrm{th}}=4m_{2}^{2} s th = 4 m 2 2 and d ∗ = ( s 1 + s 2 − 2 m 1 2 − 2 m 2 2 ) / 2 d_{*}=(s_{1}+s_{2}-2m_{1}^{2}-2m_{2}^{2})/2 d ∗ = ( s 1 + s 2 − 2 m 1 2 − 2 m 2 2 ) /2 . The coefficients in Table X and Table XVI obey
Table X. Coefficients of scalar loop integrals in the decomposition of i M 6 i\mathcal{M}_{6} i M 6
f B 6 11 ( s t h ) + f C 6 1 ( s t h ) d ∗ = 0 , f B 12 11 ( s t h ) + f C 12 1 ( s t h ) d ∗ = 0. (36) \frac{f_{B6}^{11}(s_{\mathrm{th}})+f_{C6}^{1}(s_{\mathrm{th}})}{d_{*}}=0,\qquad f_{B12}^{11}(s_{\mathrm{th}})+\frac{f_{C12}^{1}(s_{\mathrm{th}})}{d_{*}}=0. \tag*{(36)} d ∗ f B 6 11 ( s th ) + f C 6 1 ( s th ) = 0 , f B 12 11 ( s th ) + d ∗ f C 12 1 ( s th ) = 0. (36)
These identities hold for arbitrary m 1 m_{1} m 1 and hence also for the subtraction term with m 1 2 → Λ 1 2 m_{1}^{2}\to\Lambda_{1}^{2} m 1 2 → Λ 1 2 . Provided that the regulator-induced singularities lie outside this local region, terms with m 2 m_{2} m 2 or m 3 m_{3} m 3 shifted have no branch point at the physical s t h s_{\mathrm{th}} s th . The leading square-root term therefore also cancels in the complete subtracted amplitudes M 6 , 12 \mathcal{M}_{6,12} M 6 , 12 ; their normal-threshold nonanalyticity starts at order ∣ δ ∣ 3 / 2 |\delta|^{3/2} ∣ δ ∣ 3/2 or higher. This cancellation does not remove all threshold dependence.
In a separate neighborhood of a nondegenerate Landau point that excludes s t h s_{\mathrm{th}} s th , the triangle integral has the standard logarithmic singularity [1–3], which we write locally as
C 0 ( s 3 ) = C 0 r e g , L ( s 3 ) + ρ C log ( Δ + i 0 μ 2 ) + ⋯ . (37) C_{0}(s_{3})=C_{0}^{\mathrm{reg,L}}(s_{3})+\rho_{C}\log\left(\frac{\Delta+i0}{\mu^{2}}\right)+\cdots. \tag*{(37)} C 0 ( s 3 ) = C 0 reg , L ( s 3 ) + ρ C log ( μ 2 Δ + i 0 ) + ⋯ . (37)
Here ρ C \rho_C ρ C is the coefficient of the leading logarithm for the chosen scalar-integral normalization and analytic-continuation convention, and μ \mu μ is an arbitrary reference mass that makes the logarithm dimensionless. Changing μ \mu μ only shifts the locally regular part. The ordinary-threshold part of C 0 C_0 C 0 is locally regular in this neighborhood, as are the bubble integrals. Before the form-factor subtraction, the Landau-logarithmic contribution is consequently
i M i , k L , n o n a n ( s 3 ) = g i ρ C C ~ i k ( s 3 ) s 3 − M 3 2 + i M 3 Γ 3 log ( Δ + i 0 μ 2 ) + ⋯ . (38) i\mathcal{M}^{\mathrm{L,nonan}}_{i,k}(s_3)=\frac{g_i\rho_C\widetilde{C}^{k}_{i}(s_3)}{s_3-M_3^2+iM_3\Gamma_3}\log\left(\frac{\Delta+i0}{\mu^2}\right)+\cdots. \tag*{(38)} i M i , k L , nonan ( s 3 ) = s 3 − M 3 2 + i M 3 Γ 3 g i ρ C C i k ( s 3 ) log ( μ 2 Δ + i 0 ) + ⋯ . (38)
Equations (35) and (38) describe different local limits, rather than a single uniform expansion about both branch points. When several Lorentz scalars contribute to the same helicity amplitude, they must be added coherently before taking an absolute value. The ratios a i j k a_i^{jk} a i j k and b i j k b_i^{jk} b i j k used below compare the regulated contributions component by component; they are undefined when R [ f C i k C 0 ] = 0 \mathcal{R}[f_{Ci}^k C_0]=0 R [ f C i k C 0 ] = 0 and require particular care near a zero of this denominator.
Relative contributions and lineshape classification
To evaluate the regulated contributions at a common kinematic point, we specify the remaining invariant of the 1 → 3 1 \to3 1 → 3 process. At fixed initial and final particle masses, there are two independent Dalitz invariants, which we choose as s 3 = ( k 1 + k 2 ) 2 s_3=(k_1+k_2)^2 s 3 = ( k 1 + k 2 ) 2 and s 23 = ( p 2 + k 2 ) 2 s_{23}=(p_2+k_2)^2 s 23 = ( p 2 + k 2 ) 2 . At s 3 = s 3 − s_3=s_3^- s 3 = s 3 − , the allowed range of s 23 s_{23} s 23 is
s 23 min = ( E ~ k 2 + E ~ p 2 ) 2 − ( E ~ k 2 2 − k 2 2 + E ~ p 2 2 − s 2 ) 2 , s 23 max = ( E ~ k 2 + E ~ p 2 ) 2 − ( E ~ k 2 2 − k 2 2 − E ~ p 2 2 − s 2 ) 2 . (39) \begin{aligned}
s_{23}^{\min}&=(\widetilde{E}_{k_2}+\widetilde{E}_{p_2})^2-\left(\sqrt{\widetilde{E}_{k_2}^2-k_2^2}+\sqrt{\widetilde{E}_{p_2}^2-s_2}\right)^2,\\
s_{23}^{\max}&=(\widetilde{E}_{k_2}+\widetilde{E}_{p_2})^2-\left(\sqrt{\widetilde{E}_{k_2}^2-k_2^2}-\sqrt{\widetilde{E}_{p_2}^2-s_2}\right)^2 . \tag*{(39)}
\end{aligned} s 23 m i n s 23 m a x = ( E k 2 + E p 2 ) 2 − ( E k 2 2 − k 2 2 + E p 2 2 − s 2 ) 2 , = ( E k 2 + E p 2 ) 2 − ( E k 2 2 − k 2 2 − E p 2 2 − s 2 ) 2 . (39)
where
E ~ k 2 = 1 2 s 3 ( s 3 + k 2 2 − k 1 2 ) , E ~ p 2 = 1 2 s 3 ( p 1 2 − s 3 − s 2 ) . (40) \widetilde{E}_{k_2}=\frac{1}{2\sqrt{s_3}}(s_3+k_2^2-k_1^2),\qquad\widetilde{E}_{p_2}=\frac{1}{2\sqrt{s_3}}(p_1^2-s_3-s_2). \tag*{(40)} E k 2 = 2 s 3 1 ( s 3 + k 2 2 − k 1 2 ) , E p 2 = 2 s 3 1 ( p 1 2 − s 3 − s 2 ) . (40)
To numerically compute a i j k a_i^{jk} a i j k and b i j k b_i^{jk} b i j k , we take
s 23 = s 23 min + s 23 max 2 = 2 E ~ k 2 E ~ p 2 + k 2 2 + s 2 . (41) s_{23}=\frac{s_{23}^{\min}+s_{23}^{\max}}{2}=2\widetilde{E}_{k_2}\widetilde{E}_{p_2}+k_2^2+s_2 . \tag*{(41)} s 23 = 2 s 23 m i n + s 23 m a x = 2 E k 2 E p 2 + k 2 2 + s 2 . (41)
The results for a i j k a_i^{jk} a i j k , b i j k b_i^{jk} b i j k , and ∣ R [ f C i k C 0 ] ∣ |\mathcal{R}[f_{Ci}^k C_0]| ∣ R [ f C i k C 0 ] ∣ are shown in Table III .
The very small absolute C 0 C_0 C 0 contributions in Table III for the cases illustrated in Figure 2 (k) and Figure 2 (l) result from the midpoint choice of s 23 s_{23} s 23 . For k 1 2 = k 2 2 = M f 2 k_1^2=k_2^2=M_f^2 k 1 2 = k 2 2 = M f 2 , Eq. (41) gives
s 23 m i d = M f 2 + s 1 + s 2 − s 3 2 , H ≡ 2 M f 2 + s 1 + s 2 − s 3 − 2 s 23 = 0 at s 23 = s 23 m i d . (42) s_{23}^{\mathrm{mid}}=M_f^2+\frac{s_1+s_2-s_3}{2},\qquad H\equiv2M_f^2+s_1+s_2-s_3-2s_{23}=0\quad\text{at }s_{23}=s_{23}^{\mathrm{mid}} . \tag*{(42)} s 23 mid = M f 2 + 2 s 1 + s 2 − s 3 , H ≡ 2 M f 2 + s 1 + s 2 − s 3 − 2 s 23 = 0 at s 23 = s 23 mid . (42)
The coefficients of M 11 , 12 \mathcal{M}_{11,12} M 11 , 12 contain this common factor H H H , which is independent of the propagator masses and hence is preserved by the form-factor subtraction. The values 4.97 × 10 − 16 4.97\times10^{-16} 4.97 × 1 0 − 16 and 2.95 × 10 − 16 2.95\times10^{-16} 2.95 × 1 0 − 16 are numerical residuals of this kinematic zero, rather than additional dynamical suppression scales. This point-specific zero does not imply vanishing amplitudes away from the midpoint or vanishing invariant-mass spectra. The qualitative lineshape classification is based on the full amplitudes and spectra, not on these residual magnitudes, and is therefore unchanged by this special choice of s 23 s_{23} s 23 .
The separate local expansions in Eqs. (35) and (38) identify the coherent normal-threshold term and the Landau logarithm, respectively, whereas Table III compares the complete coefficient-times-integral products after the form-factor subtraction. The tabulated b i j k b_i^{jk} b i j k uses R [ f B i k B 01 ] \mathcal{R}[f_{Bi}^k B_{01}] R [ f B i k B 01 ] and R [ f C i k C 0 ] \mathcal{R}[f_{Ci}^k C_0] R [ f C i k C 0 ] , including their regular parts, and is not a ratio of the leading nonanalytic terms. A zero of the physical-mass coefficient f C i k f_{Ci}^k f C i k does not require the complete subtracted C 0 C_0 C 0 contribution to vanish: the shifted-mass terms can retain a regular contribution at the physical Landau point. Likewise, a vanishing explicit B 01 B_{01} B 01 contribution after subtraction does not remove the ordinary two-particle threshold, which can also be carried by the C 0 C_0 C 0 combination. At fixed s 1 s_1 s 1 and s 2 s_2 s 2 , the B 02 B_{02} B 02 and B 03 B_{03} B 03 terms contribute to the local smooth background, while the A 0 A_0 A 0 terms cancel exactly. The table therefore shows the relative sizes of the scalar-integral contributions at the selected kinematic point, subject to the midpoint qualification above, while their energy dependence and interference determine the full lineshape.
The three classes of lineshapes can be understood in light of Table III as follows:
(I) For serial numbers 1, 4, 5, 8, 9, and 10, the regulated C 0 C_0 C 0 term dominates over the explicit B 0 B_0 B 0 terms. All three b i 1 j k b_i^{1jk} b i 1 j k vanish for i = 1 , 4 , 9 , 10 i=1,4,9,10 i = 1 , 4 , 9 , 10 , while the largest bubble ratios for the cases shown in Figure 2 (e) and 2(h) are only 6.40 × 10 − 3 6.40 \times10^{-3} 6.40 × 1 0 − 3 and 3.74 × 10 − 2 3.74 \times10^{-2} 3.74 × 1 0 − 2 , respectively. The small bubble contributions allow the logarithmic enhancement associated with the triangle integral to remain prominent. This is consistent with the pronounced narrow TS peaks shown in Figure 2 (a), 2(d), 2(e) and 2(h) to 2(j) and Figure 3 (a), 3(d), 3(e) and 3(h) to 3(j).
(II) For serial numbers 2, 3, 7, and 11, the spectra exhibit less pronounced TS peaks, and the bubble contributions can compete more strongly with the triangle term than in class (I). For both Lorentz components of the case shown in Figure 2 (b), the B 01 B_{01} B 01 and B 03 B_{03} B 03 ratios are approximately 0.438 and 0.483, respectively; the corresponding ratios for the case shown in Figure 2 (c) are 0.301 and 0.332. For the case shown in Figure 2 (g), b 7 11 = 0 b_7^{11}=0 b 7 11 = 0 , but b 7 31 ≃ 0.268 b_7^{31} \simeq0.268 b 7 31 ≃ 0.268 gives a sizable smooth contribution. These entries indicate stronger competition between the scalar-integral terms than in class (I). The tabulated ratios for the case shown in Figure 2 (k) are b 11 11 ≃ 5.94 × 10 − 2 b_{11}^{11} \simeq5.94 \times10^{-2} b 11 11 ≃ 5.94 × 1 0 − 2 , b 11 21 ≃ 0.195 b_{11}^{21} \simeq0.195 b 11 21 ≃ 0.195 , and b 11 31 ≃ 4.11 × 10 − 2 b_{11}^{31} \simeq4.11 \times10^{-2} b 11 31 ≃ 4.11 × 1 0 − 2 ; because of the midpoint zero discussed above, these ratios of numerical residuals are not used on their own to establish its lineshape classification. The full B 01 B_{01} B 01 and C 0 C_0 C 0 terms share the ordinary-threshold cut, while the smooth B 02 B_{02} B 02 and B 03 B_{03} B 03 contributions can interfere with the triangle term and reduce the prominence of the TS peak, without requiring the Landau logarithm to disappear from every Lorentz component. This interpretation is consistent with the less pronounced peaks in Figure 2 (b), 2(c), 2(g) and 2(k) and Figure 3 (b), 3(c), 3(g) and 3(k). For generic S 23 S_{23} S 23 , the coefficient f C 2 2 f_{C_2}^2 f C 2 2 contains a Landau-zero factor. At the selected midpoint, f C 2 1 f_{C_2}^1 f C 2 1 also vanishes at the physical Landau point, so both physical-mass C 0 C_0 C 0 coefficients are zero there. These zeros must be distinguished from the finite subtracted C 0 C_0 C 0 contributions in the table and do not determine the integrated lineshape by themselves.
(III) For serial numbers 6 and 12, the distinguishing dynamical feature is the Landau-zero factor in the physical-mass C 0 C_0 C 0 coefficient, which suppresses the leading logarithmic enhancement. For the case shown in Figure 2 (f), the full subtracted C 0 C_0 C 0 contribution remains finite, with magnitude 9.48 × 10 − 2 9.48 \times10^{-2} 9.48 × 1 0 − 2 , while b 6 11 ≃ 0.438 b_6^{11} \simeq0.438 b 6 11 ≃ 0.438 and b 6 31 ≃ 0.483 b_6^{31} \simeq0.483 b 6 31 ≃ 0.483 . For the case shown in Figure 2 (l), the tabulated C 0 C_0 C 0 contribution is 2.95 × 10 − 16 2.95 \times10^{-16} 2.95 × 1 0 − 16 , and the quoted bubble ratios are 1.34 × 10 − 3 1.34 \times10^{-3} 1.34 × 1 0 − 3 , 8.13 × 10 − 4 8.13 \times10^{-4} 8.13 × 1 0 − 4 , and 4.66 × 10 − 2 4.66 \times10^{-2} 4.66 × 1 0 − 2 . Its tiny absolute contribution reflects the special S 23 S_{23} S 23 choice described above and must be distinguished from the dynamical Landau zero. Neither the ratios for the case shown in Figure 2 (f) nor the residual ratios for the case shown in Figure 2 (l) establish dominance of an explicit B 0 B_0 B 0 term. Moreover, Eq. (36) shows that the leading ordinary-threshold square roots cancel between the B 01 B_{01} B 01 and C 0 C_0 C 0 contributions in these amplitudes. The cusp-like profiles in Figure 2 (f) and 2(l) and Figure 3 (f) and 3(l) should therefore be understood through the weakened Landau enhancement, the higher-order ordinary-threshold dependence, and interference with the smooth background of the complete amplitude, rather than a large B 0 / C 0 B_0/C_0 B 0 / C 0 ratio or an isolated B 01 B_{01} B 01 square-root term.
Coefficient zeros and regulated loop contributions
We examine the zeros of all seventeen physical-mass C 0 C_{0} C 0 coefficients before the form-factor subtraction, distinguishing Landau factors from kinematic and numerator cancellations. Their initial-energy scans in Figure 4 are then compared with the B 03 B_{03} B 03 coefficients and scalar integrals in Figure 5 and Figure 6 .
Figure 4. Magnitudes of the physical-mass C 0 C_0 C 0 coefficients f C i k f_{C_i}^{k} f C i k before the form-factor subtraction, as functions of the initial-state center-of-mass energy.
Figure 5. Magnitudes of the physical-mass coefficients f B i 3 k f_{B_i}^{3k} f B i 3 k of B 03 B_{03} B 03 before the form-factor subtraction, as functions of the initial-state center-of-mass energy.
Figure 6. Evolution of ∣ B 03 ∣ |B_{03}| ∣ B 03 ∣ and ∣ C 0 ∣ |C_0| ∣ C 0 ∣ with the initial-state center-of-mass energy, evaluated at the physical masses without the form-factor subtraction.
We use the component ordering in Table II and the assumptions m 3 = m 2 m_{3}=m_{2} m 3 = m 2 and k 1 2 = k 2 2 = M f 2 k_{1}^{2}=k_{2}^{2}=M_{f}^{2} k 1 2 = k 2 2 = M f 2 of Appendix § 1.2.1 . The common variables are
d = m 1 2 − m 2 2 , U = s 1 + s 2 , V = s 1 − s 2 , Z = s 3 , G = V 2 − 2 U Z + Z 2 . (43) d=m_{1}^{2}-m_{2}^{2},\qquad U=s_{1}+s_{2},\qquad V=s_{1}-s_{2},\qquad Z=s_{3},\qquad G=V^{2}-2UZ+Z^{2}. \tag*{(43)} d = m 1 2 − m 2 2 , U = s 1 + s 2 , V = s 1 − s 2 , Z = s 3 , G = V 2 − 2 U Z + Z 2 . (43)
The factor H H H is defined in Eq. (42) . Here s 1 s_{1} s 1 , s 2 s_{2} s 2 , s 3 s_{3} s 3 , and s 23 s_{23} s 23 remain independent invariants; no midpoint condition is imposed. The Landau polynomial entering the coefficient decompositions is
P ( s 1 , s 2 , s 3 ) = m 2 4 s 3 + [ m 1 4 + s 1 s 2 − m 1 2 ( s 1 + s 2 − s 3 ) ] s 3 + m 2 2 [ ( s 1 − s 2 ) 2 − ( 2 m 1 2 + s 1 + s 2 ) s 3 ] . (44) \mathcal{P}(s_{1},s_{2},s_{3})=m_{2}^{4}s_{3}+\left[m_{1}^{4}+s_{1}s_{2}-m_{1}^{2}(s_{1}+s_{2}-s_{3})\right]s_{3}+m_{2}^{2}\left[(s_{1}-s_{2})^{2}-(2m_{1}^{2}+s_{1}+s_{2})s_{3}\right]. \tag*{(44)} P ( s 1 , s 2 , s 3 ) = m 2 4 s 3 + [ m 1 4 + s 1 s 2 − m 1 2 ( s 1 + s 2 − s 3 ) ] s 3 + m 2 2 [ ( s 1 − s 2 ) 2 − ( 2 m 1 2 + s 1 + s 2 ) s 3 ] . (44)
For m 3 = m 2 m_{3}=m_{2} m 3 = m 2 , P \mathcal{P} P equals minus four times the polynomial on the right-hand side of Eq. (4) . Its two roots in s 3 s_{3} s 3 are
s 3 ± = 4 m 2 2 + ( m 1 2 + m 2 2 − s 2 ) ( s 1 − m 1 2 − m 2 2 ) − 4 m 1 2 m 2 2 ± λ ( s 1 , m 1 2 , m 2 2 ) λ ( s 2 , m 1 2 , m 2 2 ) 2 m 1 2 , (45) s_{3}^{\pm}=4m_{2}^{2}+\frac{(m_{1}^{2}+m_{2}^{2}-s_{2})(s_{1}-m_{1}^{2}-m_{2}^{2})-4m_{1}^{2}m_{2}^{2}\pm\sqrt{\lambda(s_{1},m_{1}^{2},m_{2}^{2})\lambda(s_{2},m_{1}^{2},m_{2}^{2})}}{2m_{1}^{2}}, \tag*{(45)} s 3 ± = 4 m 2 2 + 2 m 1 2 ( m 1 2 + m 2 2 − s 2 ) ( s 1 − m 1 2 − m 2 2 ) − 4 m 1 2 m 2 2 ± λ ( s 1 , m 1 2 , m 2 2 ) λ ( s 2 , m 1 2 , m 2 2 ) , (45)
where s 3 − s_{3}^{-} s 3 − is the branch satisfying the Coleman–Norton condition in the kinematic window considered here. Thus,
P ( s 1 , s 2 , s 3 ) = m 1 2 ( s 3 − s 3 − ) ( s 3 − s 3 + ) . (46) \mathcal{P}(s_{1},s_{2},s_{3})=m_{1}^{2}(s_{3}-s_{3}^{-})(s_{3}-s_{3}^{+}). \tag*{(46)} P ( s 1 , s 2 , s 3 ) = m 1 2 ( s 3 − s 3 − ) ( s 3 − s 3 + ) . (46)
To display the other zero conditions compactly, we also define two linear factors,
Ξ = U − Z − 2 d , Ω = U − 2 Z − d + m 2 2 , (47) \Xi=U-Z-2d,\qquad\Omega=U-2Z-d+m_{2}^{2}, \tag*{(47)} Ξ = U − Z − 2 d , Ω = U − 2 Z − d + m 2 2 , (47)
and five auxiliary polynomials,
Q = Z ( s 1 + 2 s 2 − 2 Z − 3 d ) + 2 m 2 2 ( V + Z ) , (48) Q=Z(s_{1}+2s_{2}-2Z-3d)+2m_{2}^{2}(V+Z), \tag*{(48)} Q = Z ( s 1 + 2 s 2 − 2 Z − 3 d ) + 2 m 2 2 ( V + Z ) , (48)
J = − d ( V − Z ) + s 2 ( V + Z ) , (49) \mathcal{J}=-d(V-Z)+s_{2}(V+Z), \tag*{(49)} J = − d ( V − Z ) + s 2 ( V + Z ) , (49)
N = G + ( d − s 2 ) ( V + Z ) − Z ( U − Z ) , (50) \mathcal{N}=G+(d-s_{2})(V+Z)-Z(U-Z), \tag*{(50)} N = G + ( d − s 2 ) ( V + Z ) − Z ( U − Z ) , (50)
K = G Q − ( V + Z ) P , (51) \mathcal{K}=GQ-(V+Z)\mathcal{P}, \tag*{(51)} K = GQ − ( V + Z ) P , (51)
W = m 2 2 ( Z 2 − V 2 ) + Z 2 [ − 6 d V + 4 d Z + 3 U V + 9 U Z − 5 V 2 − 3 V Z − 4 Z 2 ] . (52) \mathcal{W}=m_{2}^{2}(Z^{2}-V^{2})+\frac{Z}{2}\left[-6dV+4dZ+3UV+9UZ-5V^{2}-3VZ-4Z^{2}\right]. \tag*{(52)} W = m 2 2 ( Z 2 − V 2 ) + 2 Z [ − 6 d V + 4 d Z + 3 U V + 9 U Z − 5 V 2 − 3 V Z − 4 Z 2 ] . (52)
With these definitions, the coefficients sharing N \mathcal{N} N take the forms
f C 1 1 = − N G , f C 4 1 = N G , (53) f_{C_{1}}^{1}=-\frac{\mathcal{N}}{G},\qquad f_{C_{4}}^{1}=\frac{\mathcal{N}}{G}, \tag*{(53)} f C 1 1 = − G N , f C 4 1 = G N , (53)
f C 5 1 = 2 G [ ( d + 3 V − 3 Z ) N + 2 G ( − d + 2 m 2 2 − s 1 + 2 s 2 + 2 Z ) ] . (54) f_{C_{5}}^{1}=\frac{2}{G}\left[(d+3V-3Z)\mathcal{N}+2G(-d+2m_{2}^{2}-s_{1}+2s_{2}+2Z)\right]. \tag*{(54)} f C 5 1 = G 2 [ ( d + 3 V − 3 Z ) N + 2 G ( − d + 2 m 2 2 − s 1 + 2 s 2 + 2 Z ) ] . (54)
The two components of each of f C 2 f_{C_{2}} f C 2 and f C 3 f_{C_{3}} f C 3 are
f C 2 1 = [ G + 2 H ( V + Z ) ] P − H K G 2 , (55) f_{C_{2}}^{1}=\frac{[G+2H(V+Z)]\mathcal{P}-H\mathcal{K}}{G^{2}}, \tag*{(55)} f C 2 1 = G 2 [ G + 2 H ( V + Z )] P − H K , (55)
f C 2 2 = 2 P G , (56) f_{C_{2}}^{2}=\frac{2\mathcal{P}}{G}, \tag*{(56)} f C 2 2 = G 2 P , (56)
f C 3 1 = [ G + H ( V + Z ) ] K − 8 H s 1 Z P 4 G 2 , (57) f_{C_{3}}^{1}=\frac{[G+H(V+Z)]\mathcal{K}-8Hs_{1}Z\mathcal{P}}{4G^{2}}, \tag*{(57)} f C 3 1 = 4 G 2 [ G + H ( V + Z )] K − 8 H s 1 Z P , (57)
f C 3 2 = K 2 G . (58) f_{C_{3}}^{2}=\frac{\mathcal{K}}{2G}. \tag*{(58)} f C 3 2 = 2 G K . (58)
The next two coefficients distinguish an overall Landau factor from a difference of terms:
f C 6 1 = − 4 P G . (59) f_{C_{6}}^{1}=-\frac{4\mathcal{P}}{G}. \tag*{(59)} f C 6 1 = − G 4 P . (59)
f C 7 1 = 2 ( Z Ξ Ω − 2 P ) G . (60) f_{C_7}^{1}=\frac{2(Z\Xi\Omega-2\mathcal{P})}{G}. \tag*{(60)} f C 7 1 = G 2 ( Z ΞΩ − 2 P ) . (60)
The three components of f C 8 f_{C_8} f C 8 are
f C 8 1 = P − 2 G Ω 2 G , (61) f_{C_8}^{1}=\frac{\mathcal{P}-2G\Omega}{2G}, \tag*{(61)} f C 8 1 = 2 G P − 2 G Ω , (61)
f C 8 2 = 3 ( V + Z ) P − G ( Q + 4 T ) G 2 , (62) f_{C_8}^{2}=\frac{3(V+Z)\mathcal{P}-G(Q+4\mathcal{T})}{G^2}, \tag*{(62)} f C 8 2 = G 2 3 ( V + Z ) P − G ( Q + 4 T ) , (62)
f C 8 3 = − [ G + 3 Z ( U − Z ) ] P + G W Z G 2 . (63) f_{C_8}^{3}=-\frac{[G+3Z(U-Z)]\mathcal{P}+G\mathcal{W}}{ZG^2}. \tag*{(63)} f C 8 3 = − Z G 2 [ G + 3 Z ( U − Z )] P + G W . (63)
Finally, the two components of f C 9 f_{C_9} f C 9 and the coefficients f C 10 1 f_{C_{10}}^{1} f C 10 1 , f C 11 1 f_{C_{11}}^{1} f C 11 1 , and f C 12 1 f_{C_{12}}^{1} f C 12 1 read
f C 9 1 = G − Z Ξ G , f C 9 2 = T G , (64) f_{C_9}^{1}=\frac{G-Z\Xi}{G},\qquad f_{C_9}^{2}=\frac{\mathcal{T}}{G}, \tag*{(64)} f C 9 1 = G G − Z Ξ , f C 9 2 = G T , (64)
f C 10 1 = Ω , f C 11 1 = Z H Ξ Ω G , (65) f_{C_{10}}^{1}=\Omega,\qquad f_{C_{11}}^{1}=\frac{ZH\Xi\Omega}{G}, \tag*{(65)} f C 10 1 = Ω , f C 11 1 = G Z H ΞΩ , (65)
f C 12 1 = − Z H P G . (66) f_{C_{12}}^{1}=-\frac{ZH\mathcal{P}}{G}. \tag*{(66)} f C 12 1 = − G Z H P . (66)
On the q 2 q 3 q_2q_3 q 2 q 3 cut in the rest frame of p 3 p_3 p 3 , let z z z be the cosine of the angle between p 2 p_2 p 2 and q 3 q_3 q 3 . The uncut denominator is
D 1 = a − b z , a = Ξ 2 , b = 2 ∣ p 2 ∣ ∣ q 3 ∣ = G λ ( Z , m 2 2 , m 2 2 ) 2 Z . (67) D_1=a-bz,\qquad a=\frac{\Xi}{2},\qquad b=2\lvert p_2\rvert\lvert q_3\rvert=\frac{\sqrt{G\lambda(Z,m_2^2,m_2^2)}}{2Z}. \tag*{(67)} D 1 = a − b z , a = 2 Ξ , b = 2 ∣ p 2 ∣ ∣ q 3 ∣ = 2 Z G λ ( Z , m 2 2 , m 2 2 ) . (67)
For b ≠ 0 b\ne0 b = 0 , its angular pole lies at z 0 = a / b z_0=a/b z 0 = a / b , and
P = Z ( a 2 − b 2 ) = Z b 2 ( z 0 2 − 1 ) . (68) \mathcal{P}=Z(a^2-b^2)=Zb^2(z_0^2-1). \tag*{(68)} P = Z ( a 2 − b 2 ) = Z b 2 ( z 0 2 − 1 ) . (68)
Thus, P = 0 \mathcal{P}=0 P = 0 places the pole at a collinear endpoint z 0 = ± 1 z_0=\pm1 z 0 = ± 1 . A physical TS also requires positive Feynman parameters and the Coleman–Norton configuration; it is a logarithmic branch point, not a resonance pole.
The overall factor P \mathcal{P} P in f C 2 2 f_{C_2}^{2} f C 2 2 , f C 6 1 f_{C_6}^{1} f C 6 1 , and f C 12 1 f_{C_{12}}^{1} f C 12 1 suppresses their leading Landau logarithms at generic s 23 s_{23} s 23 . The first component of f C 2 f_{C_2} f C 2 shares this zero at the midpoint:
f C 2 1 ∣ H = 0 = P G = 1 2 f C 2 2 . (69) \left.f_{C_2}^{1}\right|_{H=0}=\frac{\mathcal{P}}{G}=\frac{1}{2}f_{C_2}^{2}. \tag*{(69)} f C 2 1 H = 0 = G P = 2 1 f C 2 2 . (69)
In contrast, f C 11 1 f_{C_{11}}^{1} f C 11 1 contains the external midpoint factor H H H and the mass-dependent factors Ξ \Xi Ξ and Ω \Omega Ω , the latter shared with f C 10 1 f_{C_{10}}^{1} f C 10 1 . The common vertex contraction in the amplitudes associated with Figure 4 (j) and Figure 4 (k) satisfies
( p 2 − q 3 ) ⋅ ( p 1 + q 2 ) ∣ D 1 = D 2 = D 3 = 0 = s 1 + s 2 − 2 Z − m 1 2 + 2 m 2 2 = Ω . (70) \left.(p_2-q_3)\cdot(p_1+q_2)\right|_{D_1=D_2=D_3=0}=s_1+s_2-2Z-m_1^2+2m_2^2=\Omega. \tag*{(70)} ( p 2 − q 3 ) ⋅ ( p 1 + q 2 ) ∣ D 1 = D 2 = D 3 = 0 = s 1 + s 2 − 2 Z − m 1 2 + 2 m 2 2 = Ω. (70)
Hence Ω = 0 \Omega=0 Ω = 0 is a numerator-contraction zero. For f C 11 1 f_{C_{11}}^{1} f C 11 1 , the azimuthal average of ( q 3 − q 2 ) ⋅ ( k 1 − k 2 ) (q_3-q_2)\cdot(k_1-k_2) ( q 3 − q 2 ) ⋅ ( k 1 − k 2 ) on the two-particle cut is H ∣ q 3 ∣ z / ∣ p 2 ∣ H\lvert q_3\rvert z/\lvert p_2\rvert H ∣ q 3 ∣ z / ∣ p 2 ∣ , which becomes Z H Ξ / G ZH\Xi/G Z H Ξ/ G at z = z 0 z=z_0 z = z 0 . Thus, Ξ = 0 \Xi=0 Ξ = 0 is a longitudinal-projection zero at the angular midpoint z 0 = 0 z_0=0 z 0 = 0 , not a new endpoint singularity. Neither factor by itself establishes a physical TS.
Other zeros follow from numerator cancellations. The pair f C 1 1 , f C 4 1 f_{C_1}^{1},f_{C_4}^{1} f C 1 1 , f C 4 1 shares N = 0 N=0 N = 0 , whereas f C 5 1 f_{C_5}^{1} f C 5 1 requires cancellation between the two terms in Eq. (54) . The H H H -dependent terms likewise allow f C 2 1 f_{C_2}^{1} f C 2 1 and f C 3 1 f_{C_3}^{1} f C 3 1 to vanish away from the roots of P \mathcal{P} P and K \mathcal{K} K , respectively. Although f C 3 1 = f C 3 2 / 2 f_{C_3}^{1}=f_{C_3}^{2}/2 f C 3 1 = f C 3 2 /2 at H = 0 H=0 H = 0 , their zeros generally differ because
f C 3 1 ∣ K = 0 = − 2 H s 1 Z P G 2 , (71) \left.f_{C_3}^{1}\right|_{\mathcal{K}=0}=-\frac{2Hs_1Z\mathcal{P}}{G^2}, \tag*{(71)} f C 3 1 K = 0 = − G 2 2 H s 1 Z P , (71)
which need not vanish. The distinct cancellation conditions for f C 7 1 f_{C_7}^{1} f C 7 1 , f C 8 1 , 2 , 3 f_{C_8}^{1,2,3} f C 8 1 , 2 , 3 , and f C 9 1 , 2 f_{C_9}^{1,2} f C 9 1 , 2 follow directly from Eqs. (60) to (64) . These expressions assume G ≠ 0 G\ne0 G = 0 ; at Gram degeneracy the complete helicity amplitude must be assembled before taking the limit. The apparent denominator factor Z Z Z in f C 8 3 f_{C_8}^{3} f C 8 3 is removable.
At fixed s 1 s_1 s 1 , s 2 s_2 s 2 , the overall Landau factor behaves near Δ = s 3 − s 3 + = 0 \Delta=s_3-s_3^+=0 Δ = s 3 − s 3 + = 0 as
P = m 1 2 ( s 3 − s 3 + ) Δ + O ( Δ 2 ) . (72) \mathcal{P}=m_1^2(s_3-s_3^+)\Delta+O(\Delta^2). \tag*{(72)} P = m 1 2 ( s 3 − s 3 + ) Δ + O ( Δ 2 ) . (72)
With smooth nonzero prefactors, it changes log [ ( Δ + i 0 ) / μ 2 ] \log[(\Delta+i0)/\mu^2] log [( Δ + i 0 ) / μ 2 ] into Δ log [ ( Δ + i 0 ) / μ 2 ] \Delta\log[(\Delta+i0)/\mu^2] Δ log [( Δ + i 0 ) / μ 2 ] . An additional H H H zero can increase the suppression along a special path, but neither cancellation creates a new pole nor necessarily removes all nonanalytic dependence.
The scan in Figure 4 varies s 1 s_1 s 1 at fixed s 2 s_2 s 2 , s 3 s_3 s 3 , s 23 s_{23} s 23 , using m 1 = 0.891 m_1=0.891 m 1 = 0.891 GeV, m 2 = m 3 = 0.493 m_2=m_3=0.493 m 2 = m 3 = 0.493 GeV, s 2 = 0.135 \sqrt{s_2}=0.135 s 2 = 0.135 GeV, and M f = 0.139 M_f=0.139 M f = 0.139 GeV. At s 1 ∗ = ( 1.42 G e V ) 2 s_{1*}=(1.42\ \mathrm{GeV})^2 s 1 ∗ = ( 1.42 GeV ) 2 , Eq. (5) gives s 3 ∗ ≃ 0.973861381 G e V 2 s_{3*}\simeq0.973861381\ \mathrm{GeV}^2 s 3 ∗ ≃ 0.973861381 GeV 2 and the midpoint is s 23 ∗ ≃ 0.549702810 G e V 2 s_{23*}\simeq0.549702810\ \mathrm{GeV}^2 s 23 ∗ ≃ 0.549702810 GeV 2 . Holding these values fixed yields
s 23 = s 23 ∗ = M f 2 + s 1 ∗ + s 2 − s 3 ∗ 2 , H = s 1 − s 1 ∗ . (73) s_{23}=s_{23*}=M_f^2+\frac{s_{1*}+s_2-s_{3*}}{2},\qquad H=s_1-s_{1*}. \tag*{(73)} s 23 = s 23 ∗ = M f 2 + 2 s 1 ∗ + s 2 − s 3 ∗ , H = s 1 − s 1 ∗ . (73)
Thus, the scan does not follow the moving midpoint: changing s 23 ∗ s_{23*} s 23 ∗ moves the H H H zero but not the Landau singularity. Roots are evaluated using the exact Landau definition of s 3 ∗ s_{3*} s 3 ∗ ; G , Z > 0 G,Z>0 G , Z > 0 throughout the plotted interval. This s 1 s_1 s 1 scan is distinct from the s 3 s_3 s 3 spectra in Figure 2 and Figure 3 .
The normal q 1 q 2 q_1q_2 q 1 q 2 threshold shared by B 03 \mathbf{B}_{03} B 03 and C 0 C_0 C 0 is s 1 , t h = m 1 + m 2 = 1.384 \sqrt{s_{1,\mathrm{th}}}=m_1+m_2=1.384 s 1 , th = m 1 + m 2 = 1.384 GeV, below the scan. The q 2 q 3 q_2q_3 q 2 q 3 threshold at s 3 , t h = m 2 + m 3 = 0.986 \sqrt{s_{3,\mathrm{th}}}=m_2+m_3=0.986 s 3 , th = m 2 + m 3 = 0.986 GeV belongs to the fixed s 3 s_3 s 3 channel. These reduced two-propagator Landau branch points are not crossed here; none of the listed coefficient zeros is a normal-threshold opening.
Along this path, the Landau polynomial and the two linear factors become
P = m 2 2 ( s 1 − s 1 ∗ ) ( s 1 − s 1 † ) , s 1 † = 2 s 2 + ( m 1 2 + m 2 2 − s 2 ) s 3 ∗ m 2 2 − s 1 ∗ , (74) \mathcal{P}=m_2^2(s_1-s_{1*})(s_1-s_{1\dagger}),\qquad s_{1\dagger}=2s_2+\frac{(m_1^2+m_2^2-s_2)s_{3*}}{m_2^2}-s_{1*}, \tag*{(74)} P = m 2 2 ( s 1 − s 1 ∗ ) ( s 1 − s 1 † ) , s 1 † = 2 s 2 + m 2 2 ( m 1 2 + m 2 2 − s 2 ) s 3 ∗ − s 1 ∗ , (74)
Ξ = s 1 − r Ξ , r Ξ = s 3 ∗ + 2 d − s 2 , Ω = s 1 − r Ω , r Ω = 2 s 3 ∗ + d − m 2 2 − s 2 . (75) \Xi=s_1-r_{\Xi},\qquad r_{\Xi}=s_{3*}+2d-s_2,\qquad\Omega=s_1-r_{\Omega},\qquad r_{\Omega}=2s_{3*}+d-m_2^2-s_2. \tag*{(75)} Ξ = s 1 − r Ξ , r Ξ = s 3 ∗ + 2 d − s 2 , Ω = s 1 − r Ω , r Ω = 2 s 3 ∗ + d − m 2 2 − s 2 . (75)
At s 1 = 1.42 \sqrt{s_1}=1.42 s 1 = 1.42 GeV, the simple zeros of P \mathcal{P} P and H H H give simple zeros of f C 2 2 , f C 6 1 f_{C_2}^2,f_{C_6}^1 f C 2 2 , f C 6 1 but a double zero of f C 12 1 f_{C_{12}}^1 f C 12 1 . The latter is a coincidence of numerator factors, not two Landau branch points. Their second common zero at s 1 † ≃ 1.449776 \sqrt{s_{1\dagger}}\simeq1.449776 s 1 † ≃ 1.449776 GeV fails the Coleman–Norton condition and produces no physical TS enhancement.
The first component of f C 2 2 f_{C_2}^2 f C 2 2 also shares the reference-point zero:
f C 2 1 = s 1 − s 1 ∗ G 2 { m 2 2 ( s 1 − s 1 † ) [ G + 2 ( s 1 − s 1 ∗ ) ( V + Z ) ] − K } . (76) f_{C_2}^1=\frac{s_1-s_{1*}}{G^2}\left\{m_2^2(s_1-s_{1\dagger})[G+2(s_1-s_{1*})(V+Z)]-\mathcal{K}\right\}. \tag*{(76)} f C 2 1 = G 2 s 1 − s 1 ∗ { m 2 2 ( s 1 − s 1 † ) [ G + 2 ( s 1 − s 1 ∗ ) ( V + Z )] − K } . (76)
The cubic in braces has one root in the interval, at s 1 ≃ 1.430925 \sqrt{s_1}\simeq1.430925 s 1 ≃ 1.430925 GeV, distinct from the second root of f C 2 2 f_{C_2}^2 f C 2 2 . The linear factors give
f C 10 1 = s 1 − r Ω , f C 11 1 = s 3 ∗ G ( s 1 − s 1 ∗ ) ( s 1 − r Ξ ) ( s 1 − r Ω ) . (77) f_{C_{10}}^1=s_1-r_{\Omega},\qquad f_{C_{11}}^1=\frac{s_{3*}}{G}(s_1-s_{1*})(s_1-r_{\Xi})(s_1-r_{\Omega}). \tag*{(77)} f C 10 1 = s 1 − r Ω , f C 11 1 = G s 3 ∗ ( s 1 − s 1 ∗ ) ( s 1 − r Ξ ) ( s 1 − r Ω ) . (77)
The three roots of f C 11 1 f_{C_{11}}^1 f C 11 1 arise from H , Ξ , Ω H,\Xi,\Omega H , Ξ , Ω ; the last is shared with f C 10 1 f_{C_{10}}^1 f C 10 1 . Neither the Ξ \Xi Ξ nor the Ω \Omega Ω root is a Landau root for these inputs.
After cancelling common numerator and denominator factors, all scanned coefficients can be written as
f C i k ∣ s c a n = R i k ( s 1 ) D i k ( s 1 ) ∏ a ( s 1 − r i k , a ) n i k , a , 1.4 G e V ≤ s 1 ≤ 1.5 G e V . (78) f_{C_i}^{k}\big|_{\mathrm{scan}}=\frac{R_{ik}(s_1)}{D_{ik}(s_1)}\prod_a(s_1-r_{ik,a})^{n_{ik,a}},\qquad1.4\ \mathrm{GeV}\le\sqrt{s_1}\le1.5\ \mathrm{GeV}. \tag*{(78)} f C i k scan = D ik ( s 1 ) R ik ( s 1 ) a ∏ ( s 1 − r ik , a ) n ik , a , 1.4 GeV ≤ s 1 ≤ 1.5 GeV . (78)
Here D i k D_{ik} D ik is the reduced denominator and R i k R_{ik} R ik has no zeros in the interval. All roots r i k , a r_{ik,a} r ik , a , their defining conditions, and multiplicities are summarized in Table IV, with energies rounded to the displayed precision.
The other roots in Table IV are numerator cancellations, not new singularities. The f C 3 1 , 2 f_{C_3}^{1,2} f C 3 1 , 2 roots differ by approximately 0.0557 MeV despite appearing coincident in Figure 4 (c). Likewise, the f C 8 1 f_{C_8}^1 f C 8 1 root differs from the nearby Ω \Omega Ω root, since f C 8 1 ∣ Ω = 0 = P / ( 2 G ) ≠ 0 f_{C_8}^1|_{\Omega=0}=\mathcal{P}/(2G)\ne0 f C 8 1 ∣ Ω = 0 = P / ( 2 G ) = 0 . The table also identifies the components with no root in the interval.
Since s 1 − r = ( E − r ) ( E + r ) s_1-r=(E-\sqrt{r})(E+\sqrt{r}) s 1 − r = ( E − r ) ( E + r ) with E = s 1 > 0 E=\sqrt{s_1}>0 E = s 1 > 0 , a root of multiplicity n n n gives ∣ f C i k ∣ ∝ ∣ E − E 0 ∣ n |f_{C_i}^k|\propto|E-E_0|^n ∣ f C i k ∣ ∝ ∣ E − E 0 ∣ n locally. Finite sampling on a logarithmic plot turns an exact zero into a finite-depth minimum, whose depth has no independent physical significance.
The B 03 B_{03} B 03 scan in Figure 5 distinguishes kinematic zeros from subtraction cancellations. The mass-independent coefficients in Figure 5 (a), Figure 5 (d), Figure 5 (i) and Figure 5 (j) remain nonzero, but their complete bubble products vanish under subtraction of the missing propagator mass, explaining the zero b i 3 k b_i^{3k} b i 3 k entries in Table III . By contrast, f B 11 3 f_{B_{11}}^{3} f B 11 3 and f B 12 3 f_{B_{12}}^{3} f B 12 3 vanish at the reference point through H = 0 H=0 H = 0 .
In Figure 6 , B 03 \mathbf{B}_{03} B 03 varies smoothly, whereas C 0 C_0 C 0 is enhanced at the physical TS near 1.42 GeV but not at the second Landau root. Their different mass dimensions preclude interpreting the plotted magnitudes as relative amplitude weights.
Coefficient zeros likewise do not determine zeros of the complete regulated amplitude. Individual component zeros can depend on the Lorentz basis, and all contributions must be combined coherently. Provided regulator-induced singularities remain outside the physical TS neighborhood, shifted-mass terms can leave a regular contribution without restoring a missing leading logarithm. The normal-threshold strength instead depends on the coherent B 0 B_0 B 0 –C 0 C_0 C 0 combination in Eq. (35) ; the cancellations in Eq. (36) remove its leading square root, not all threshold dependence.
These distinctions support the preceding lineshape classification: the class-(I) coefficients remain nonzero at the reference point; those in Figure 4 (c) and Figure 4 (g) also remain nonzero, so their weaker peaks require interference rather than removal of the Landau logarithm. The midpoint-dependent zeros in Figure 4 (b) and Figure 4 (k) differ from the generic Landau factors in Figure 4 (f) and Figure 4 (l). The full s 3 s_3 s 3 -dependent amplitudes, not the coefficient minima alone, determine the spectra.
Table 4. Zeros of all seventeen physical-mass C 0 C_0 C 0 coefficients along the fixed-s 3 s_3 s 3 , fixed-s 23 s_{23} s 23 scan. Multiplicities follow the order of the listed energies. The last row groups the components with no real zero in the plotted interval.
Summary
In this work, we have systematically investigated how the dynamical couplings at the three vertices of a triangle diagram affect the visibility and lineshape of the triangle singularity. We enumerated twelve combinations of external and internal spin-parity quantum numbers, covering seven types of effective vertices: P P S PPS P P S , V P P VPP V P P , V V P VVP V V P , V P A VPA V P A , V V V VVV V V V , V V A VVA V V A , and V V S VVS V V S . Using the Passarino–Veltman reduction, we expressed each loop amplitude as a linear combination of the scalar one-loop integrals A 0 A_0 A 0 , B 0 B_0 B 0 , and C 0 C_0 C 0 , with the corresponding Lorentz-scalar coefficients encoding the dynamical information of the vertices under consideration.
For the numerical comparison, we used the masses and kinematics of the η ( 1405 ) → π + π − π 0 \eta(1405)\to\pi^+\pi^-\pi^0 η ( 1405 ) → π + π − π 0 system with a K ∗ K ˉ K K^*\bar{K}K K ∗ K ˉ K loop, together with the f 0 ( 980 ) f_0(980) f 0 ( 980 ) resonance parameters, as a common benchmark for the twelve spin-parity assignments. We evaluated the relative weights a i j k a_i^{jk} a i j k and b i j k b_i^{jk} b i j k at the Landau singularity. The twelve cases can be classified into three categories. In the first, the C 0 C_0 C 0 contribution dominates and produces a pronounced narrow TS peak. In the second, the bubble contributions are more competitive, and their interference with the triangle term can reduce the prominence of the peak. In the third, the physical-mass coefficient of the leading Landau logarithm vanishes as a consequence of the vertex dynamics. The cusp-like spectrum is then shaped by the remaining ordinary-threshold dependence, smooth background, and coherent interference, without requiring an explicit B 0 B_0 B 0 term to become dominant. In particular, the leading normal-threshold square-root terms cancel in the complete amplitudes for cases 6 and 12, so their threshold behavior must be determined beyond that order.
These results demonstrate that the Landau condition determines the kinematic location at which a triangle singularity may occur, but does not by itself determine its prominence or detailed lineshape in the physical spectrum. The partial-wave and tensor structures of the vertices also play a crucial role. Within the effective-Lagrangian and intermediate-meson-loop framework adopted in this work, comparing the relative contributions of B 0 B_0 B 0 and C 0 C_0 C 0 allows us to analyze how the vertex dynamics affects TS interpretations of charmoniumlike and pentaquark candidates. Future TS analyses should therefore examine not only the Landau condition but also the relative magnitudes of the relevant scalar integrals and their coefficients. More precise measurements of the corresponding lineshapes at BESIII, LHCb, and Belle II will help distinguish triangle-singularity effects from contributions associated with genuine resonances.
Acknowledgments
This work is supported by the National Natural Science Foundation of China (Grant No. 12235018).
Coefficients in the scalar-integral decomposition
In this appendix, we set M k 1 = M k 2 = M f M_{k_1}=M_{k_2}=M_f M k 1 = M k 2 = M f and m 3 = m 2 m_3=m_2 m 3 = m 2 . The latter equality implies
A 02 = A 03 . (79) A_{02}=A_{03}. \tag*{(79)} A 02 = A 03 . (79)
For convenience, we use s 1 = p 1 2 s_1=p_1^2 s 1 = p 1 2 , s 2 = p 2 2 s_2=p_2^2 s 2 = p 2 2 , s 3 = ( k 1 + k 2 ) 2 s_3=(k_1+k_2)^2 s 3 = ( k 1 + k 2 ) 2 , and s 23 = ( k 2 + p 2 ) 2 s_{23}=(k_2+p_2)^2 s 23 = ( k 2 + p 2 ) 2 .
References
Table 5. Coefficients of scalar loop integrals in the decomposition of i M 1 i\mathcal{M}_{1} i M 1
Table VI. Coefficients of scalar loop integrals in the decomposition of i M 2 i\mathcal{M}_{2} i M 2
Table VII. Coefficients of scalar loop integrals in the decomposition of i M 3 i\mathcal{M}_3 i M 3
Table VII (PDF p. 20).
Table VIII. Coefficients of scalar loop integrals in the decomposition of i M 4 i\mathcal{M}_4 i M 4
Table IX. Coefficients of scalar loop integrals in the decomposition of i M 5 i\mathcal{M}_{5} i M 5
Table XI. Coefficients of scalar loop integrals in the decomposition of i M 7 i\mathcal{M}_{7} i M 7
Table XII. Coefficients of scalar loop integrals in the decomposition of i M 8 i\mathcal{M}_{8} i M 8
Table XIII. Coefficients of scalar loop integrals in the decomposition of i M 9 i\mathcal{M}_{9} i M 9
Table XIV. TABLE XIV: Coefficients of scalar loop integrals in the decomposition of i M 10 i\mathcal{M}_{10} i M 10
Table XV. TABLE XV: Coefficients of scalar loop integrals in the decomposition of i M 11 i\mathcal{M}_{11} i M 11