1 Introduction

The two-body \(D_s^+\rightarrow PP,PV\) decays with P(V) denoting the strangeless pesudoscalar (vector) meson have no configurations from the W-boson emission processes because \({{\bar{s}}}\) in \(D_s^+\) cannot be eliminated, as drawn in Fig. 1 for the topological diagrams T and C. Interestingly, it leads to a specific exploration for the annihilation mechanism applied to \(D_s^+\rightarrow \pi ^+\pi ^0,\pi ^{+}\rho ^{0},\pi ^+\omega \) [1,2,3,4,5,6].

In the short-distance W-boson annihilation (WA) \(D_s^+(c{{\bar{s}}})\rightarrow W^+\rightarrow u{{\bar{d}}}\) process, \(u{{\bar{d}}}\) can be seen to move in the opposite directions in the \(D_s^+\) rest frame, such that there exists no orbital angular momentum between them. It indicates that \(G(u{{\bar{d}}})=G(\pi ^+)=+1\) with G denoting the G-parity symmetry [3]. Since G-parity is a multiplicative quantum number, one obtains \(G(\pi ^+\rho ^0,\pi ^+\pi ^0)=(+1,-1)\). Consequently, the WA \(D_s^+\rightarrow \pi ^+\rho ^0(\pi ^+\pi ^0)\) decay due to \(G(u{{\bar{d}}})=G(\pi ^+\rho ^0)[=-G(\pi ^+\pi ^0)]\) is a G-parity conserved (violated) process, which corresponds to the experimental result \({{\mathcal {B}}}(D_s^+\rightarrow \pi ^+\rho ^0)=(1.9\pm 1.2)\times 10^{-4}\) \([{{\mathcal {B}}}(D_s^+\rightarrow \pi ^+\pi ^0)<3.4\times 10^{-4}]\) [7]. By contrast, although the WA \(D_s^+\rightarrow \pi ^+\omega \) decay violates the G-parity symmetry, \({{\mathcal {B}}}(D_s^+\rightarrow \pi ^+\omega )=(1.9\pm 0.3)\times 10^{-3}\) shows no suppression [7]. It is hence considered to receive the long-distance annihilation contribution [1, 3].

The \(D_s^+\rightarrow SP,SV\) decays can help to investigate the short and long-distance annihilation mechanisms [8,9,10], where S stands for a non-strange scalar meson. For example, the WA process for \(D_s^+\rightarrow a_0^{0(+)}\pi ^{+(0)}\) violates G-parity [9, 11], such that its branching fraction is expected as small as \({{\mathcal {B}}}(D_s^+\rightarrow \pi ^+\pi ^0)\). Nonetheless, one measures that \({{\mathcal {B}}}(D_s^+\rightarrow a_0^{0(+)}\pi ^{+(0)})\simeq 100 {{\mathcal {B}}}(D_s^+\rightarrow \pi ^+\rho ^0)\) [7, 12]. Clearly, it indicates the main contribution from the long-distance annihilation process [9]. Explicitly, the long-distance annihilation process for \(D_s^+\rightarrow a_0^{0(+)}\pi ^{+(0)}\) starts with the \(D_s^+\rightarrow \eta ^{(\prime )}\rho ^+\) weak decay, followed by the \(\eta ^{(\prime )}\) and \(\rho ^+\) rescattering. With the \(\pi ^{+(0)}\) exchange, \(\eta ^{(\prime )}\) and \(\pi ^+\) are turned into \(a_0^{+(0)}\) and \(\pi ^{0(+)}\), respectively. Since BESIII has recently reported the first observation of the branching fractions of \(D_s^+\rightarrow SV,S\rightarrow PP\) as [13, 14]

Fig. 1
figure 1

Topological diagrams for the Cabibbo-allowed \(D_s^+\) weak decays

$$\begin{aligned} {{\mathcal {B}}}_+(D_s^+\rightarrow a_0^+ \rho ^0,a_0^+\rightarrow \pi ^+\eta )&=(2.1\pm 0.8 \pm 0.5)\times 10^{-3} ,\nonumber \\ {{\mathcal {B}}}_0(D_s^+\rightarrow a_0^0 \rho ^+,a_0^0\rightarrow K^+ K^-)&=(0.7\pm 0.2\pm 0.1)\times 10^{-3} , \end{aligned}$$
(1)

we are wondering which of the short and long-distance annihilation processes can be the dominant contribution. Hence, we propose to study \(D_s^+\rightarrow a_0^{+(0)} \rho ^{0(+)}\), \(D_s^+\rightarrow a_0^+\omega \), and the resonant three-body \(D_s^+\rightarrow a_0^{+(0)} \rho ^{0(+)}, a_0^{+(0)}\rightarrow [\eta \pi ^{+(0)},K^+{{\bar{K}}}^0(K^+ K^-)]\) decays, in order to analyze the data in Eq. (1). We will also test if \({{\mathcal {B}}}(D_s^+\rightarrow a_0^+ \rho ^0, a_0^+ \rho ^0)\) have nearly equal sizes as \({{\mathcal {B}}}(D_s^+\rightarrow a_0^+ \pi ^0)\simeq {{\mathcal {B}}}(D_s^+\rightarrow a_0^0 \pi ^+)\) that respects the isospin symmetry.

2 Formalism

Fig. 2
figure 2

Triangle rescattering diagrams for \(D^+_s\rightarrow \rho ^{0(+)}a_0^{+(0)}\) and \(D^+_s\rightarrow \omega a_0^+\)

Considering the short-distance WA processes, \(D_s^+\rightarrow PP(PV)\) and \(D_s^+\rightarrow SP(SV)\) both get an A term as the annihilation amplitude in Fig. 1. According to \({{\mathcal {B}}}(D_s^+\rightarrow \pi ^+\rho ^0)\simeq 10^{-4}\) that receives the short-distance WA contribution [7], one regards A to give \({{\mathcal {B}}}(D_s^+\rightarrow a_0^{+(0)} \rho ^{0(+)})\) not larger than \(10^{-4}\). However, \(\mathcal{B}_+\) in Eq. (1) suggests \({{\mathcal {B}}}(D_s^+\rightarrow a_0^+ \rho ^0)\simeq 10^{-3}\). This strongly suggests that the main contribution to \(D_s^+\rightarrow a_0^{+(0)} \rho ^{0(+)}\) is from the long-distance annihilation process. For \(D_s^+\rightarrow a_0^+\omega \), the WA contribution is suppressed with the G-parity violation, such that \(A\simeq 0\). Therefore, we start with the triangle rescattering processes for \(D_s^+\rightarrow a_0^{+(0)} \rho ^{0(+)}\) and \(D_s^+\rightarrow a_0^+\omega \) as the most possible main contributions.

See Fig. 2, the rescattering processes for \(D_s^+\rightarrow a_0^{+(0)} \rho ^{0(+)}\) include both the weak and strong decays. In our case, the weak decays come from \(D^{+}_{s}\rightarrow \pi ^{+}\eta ^{(\prime )},K^{+}{\bar{K}}^0\), and the amplitudes are given by [2,3,4, 6]

$$\begin{aligned} {{\mathcal {M}}}_\eta (D^{+}_{s}\rightarrow \pi ^{+}\eta )= & {} \frac{G_F}{\sqrt{2}} V^*_{cs}V_{ud}(\sqrt{2}A\cos \phi -T\sin \phi ) ,\nonumber \\ {{\mathcal {M}}}_{\eta '}(D^{+}_{s}\rightarrow \pi ^{+}\eta ^{\prime })= & {} \frac{G_F}{\sqrt{2}}V^*_{cs}V_{ud}(\sqrt{2}A\sin \phi +T\cos \phi ) ,\nonumber \\ {{\mathcal {M}}}_K(D^{+}_{s}\rightarrow K^{+}{\bar{K}}^0)= & {} \frac{G_F}{\sqrt{2}}V^*_{cs}V_{ud}(C+A) , \end{aligned}$$
(2)

where \(G_F\) is the Fermi constant, \(V^*_{cs}V_{ud}\) the Cabibbo–Kobayashi–Maskawa (CKM) matrix elements, and \(V^*_{cs}V_{ud}\simeq 1\) presents the Cabibbo-allowed decay modes. In addition, (TCA) are the topological amplitudes, along with the mixing angle \(\phi =43.5^\circ \) from the \(\eta -\eta ^\prime \) mixing matrix [15, 16]:

$$\begin{aligned} \left( \begin{array}{c} \eta \\ \eta ^\prime \end{array}\right) = \left( \begin{array}{cc} \cos \phi &{} -\sin \phi \\ \sin \phi &{} \cos \phi \end{array}\right) \left( \begin{array}{c} \sqrt{1/2}(u{{\bar{u}}}+d\bar{d}) \\ s{{\bar{s}}} \end{array}\right) . \end{aligned}$$
(3)

For the strong decays \(V,S\rightarrow PP\), the amplitudes are given by [9]

$$\begin{aligned}&{{\mathcal {M}}}_{\rho ^{0(+)}\rightarrow \pi ^{+(0)}\pi ^{-(+)}}=g_\rho \epsilon \cdot (q_1-q_2),\,\nonumber \\&{{\mathcal {M}}}_{a_0^+\rightarrow \eta ^{(\prime )}\pi ^+,K^+\bar{K}^0}=g_{\eta ^{(\prime )}},g_K , \end{aligned}$$
(4)

where \(\epsilon _{\mu }\) is the polarization four vector of the \(\rho \) meson, and \(q_{1,2}^\mu \) are the four momenta of \(\pi ^+\pi ^-(\pi ^0\pi ^+)\), respectively. The SU(3) flavor symmetry is able to relate different \(V(S)\rightarrow PP\) decay channels [17, 18], such that we obtain \(g_\rho /2\) for \({{\mathcal {M}}}_{\rho ^0\rightarrow K^+ K^-,{{\bar{K}}}^0 K^0}\), \( g_\rho /\sqrt{2}\) for \({{\mathcal {M}}}_{\rho ^+\rightarrow K^+{{\bar{K}}}^0}\), and \((-)g_K/\sqrt{2} \) for \({{\mathcal {M}}}_{a^0_0\rightarrow K^+ K^-({{\bar{K}}}^0 K^0)}\), together with \((-)g_\rho /2\) for \({{\mathcal {M}}}_{\omega \rightarrow K^+ K^-(\bar{K}^0 K^0)}\).

By assembling the weak and strong couplings in the rescattering processes, we derive that

$$\begin{aligned}&{{\mathcal {M}}}(D_{s}^{+}\rightarrow \rho ^{0(+)}a^{+(0)}_{0}) ={{\mathcal {M}}}_a+\mathcal{M}_{a}^{\prime }+{{\mathcal {M}}}_{b}+{{\mathcal {M}}}_{c} , \nonumber \\&{{\mathcal {M}}}(D_{s}^{+}\rightarrow \omega a^{+}_{0}) = \hat{{\mathcal {M}}}_{b}+ \hat{{\mathcal {M}}}_{c} , \end{aligned}$$
(5)

with \(M_a^{(\prime )}\) and \(M_{b(c)}[{\hat{M}}_{b(c)}]\) for Fig. 2a, b(c), respectively. More explicitly, \({{\mathcal {M}}}_{a,b,c}\) are given by [9, 19, 20]

$$\begin{aligned}&{{\mathcal {M}}}_{a}=\int \frac{d^4{q}_{1}}{(2\pi )^{4}} \frac{\mathcal{M}_{\eta }{{\mathcal {M}}}_{\rho ^{0(+)}\rightarrow \pi ^+\pi ^{-(0)}} \mathcal{M}_{a^{+(0)}_{0}\rightarrow \eta \pi ^{+(0)}} F_{\pi }(q_{2}^2)}{(q_{1}^{2}-m_{\pi }^{2}+i\epsilon )[(q_1-p_2)^{2}-m_{\pi }^{2}+i\epsilon ][(q_1-p_1)^{2}-m_{\eta }^{2}+i\epsilon ]} ,\nonumber \\&{{\mathcal {M}}}_{b}=\int \frac{d^4{q}_{1}}{(2\pi )^{4}} \frac{{{\mathcal {M}}}_{K} {{\mathcal {M}}}_{\rho ^{0(+)}\rightarrow K^+K^-(K^+{{\bar{K}}}^0)} \mathcal{M}_{a_0^{+(0)}\rightarrow {\bar{K}}^0 K^{+(0)}} F_{K}(q_{2}^2)}{(q_{1}^{2}-m_{K}^{2}+i\epsilon )[(q_1-p_2)^{2}-m_{K}^{2}+i\epsilon ][(q_1-p_1)^{2}-m_{K}^{2}+i\epsilon ]} , \nonumber \\&{{\mathcal {M}}}_{c}=\int \frac{d^4{q}_{1}}{(2\pi )^{4}} \frac{{{\mathcal {M}}}_K {{\mathcal {M}}}_{\rho ^{0(+)}\rightarrow {\bar{K}}^0 K^{0(+)}} \mathcal{M}_{a_0^{+(0)}\rightarrow {{\bar{K}}}^0 K^{+(0)}} F_{K}(q_{2}^2)}{(q_{1}^{2}-m_{K}^{2}+i\epsilon )[(q_1-p_2)^{2}-m_{K}^{2}+i\epsilon ][(q_1-p_1)^{2}-m_{K}^{2}+i\epsilon ]} , \end{aligned}$$
(6)

with \(q_2=q_1-p_2\) and \(q_3=q_1-p_1\) following the momentum flows in Fig. 2. The form factor \(F_M(q_{2}^2)\equiv (\Lambda _M^{2}-m^{2}_M)/(\Lambda _M^{2}-q^{2}_{2})\) with the cutoff parameter \(\Lambda _M\) [\(M=(\pi ,K)\)] is to avoid the overestimation with \(q_2\) to \(\pm \infty \) [21]. Substituting \(\eta ^\prime \) and \(\omega \) for \(\eta \) in \({{\mathcal {M}}}_a\) and \(\rho ^0\) in \({{\mathcal {M}}}_{b(c)}\) leads to \({{\mathcal {M}}}_a^\prime \) and \(\hat{{\mathcal {M}}}_{b(c)}\), respectively.

As a consequence, we find that

$$\begin{aligned} {{\mathcal {M}}}(D_{s}^{+}\rightarrow \rho ^{0}a^{+}_{0})= & {} {{\mathcal {M}}}(D_{s}^{+}\rightarrow \rho ^{+}a^{0}_{0}) ,\nonumber \\ {{\mathcal {M}}}(D_{s}^{+}\rightarrow \omega a^{+}_{0})= & {} 0 , \end{aligned}$$
(7)

where the first relation respects the isospin symmetry, whereas \(\hat{{\mathcal {M}}}_{b}=- \hat{{\mathcal {M}}}_{c}\) due to \({{\mathcal {M}}}_{\omega \rightarrow K^+ K^-}=-{{\mathcal {M}}}_{\omega \rightarrow {{\bar{K}}}^0 K^0}\) cancels the rescattering contributions to \(D_{s}^{+}\rightarrow \omega a^{+}_{0}\), which causes \({{\mathcal {M}}}(D_{s}^{+}\rightarrow \omega a^{+}_{0})=0\).

To deal with the triangle loops in Eq. (6), the equation in Refs. [22,23,24,25] can be useful, given by

$$\begin{aligned}&\int \frac{d^{4}q_1}{i \pi ^2}\frac{q_1^{\mu }}{(q_1^{2}-m_{1}^{2}+i\epsilon )[(q_1-p_2)^{2}-m_{2}^{2}+i\epsilon ][(q_1-p_1)^{2}-m_{3}^{2}+i\epsilon ]}\nonumber \\&\quad = -p_\rho ^\mu C_1(p_{a_0}^2,m_\rho ^2,m_{D_s}^2,m_1^2,m_2^2,m_3^2) -p_{D_s}^\mu C_2(p_{a_0}^2,m_\rho ^2,m_{D_s}^2,m_1^2,m_2^2,m_3^2) . \end{aligned}$$
(8)

By calculating the triangle rescattering processes, we obtain

$$\begin{aligned}&\Gamma (D_{s}^{+}\rightarrow \rho ^{0{+}}a^{+(0)}_{0})= \frac{|\mathbf {p}_{2}|^{3}}{8\pi m_{\rho }^2}|H|^2 ,\nonumber \\&\quad H=\frac{1}{8\pi ^2}\bigg (g_\rho g_\eta {{\mathcal {M}}}_\eta C_\eta +g_\rho g_{\eta '}{{\mathcal {M}}}_{\eta '} C_{\eta '} +g_\rho g_K {{\mathcal {M}}}_K C_K\bigg ) , \end{aligned}$$
(9)

with \(C_{\eta ^{(\prime )}}=C_{2,\eta ^{(\prime )}}-C_{2,\eta ^{(\prime )}}^*\) and \(C_K=C_{2,K}-C_{2,K}^*\) as the integrated results of the \(C_2\) terms in Eq. (8), where \(C_{2,M}^{(*)}\) are defined by

$$\begin{aligned} C_{2,\eta }^{(*)}= & {} C_2(m_\pi ,m_\pi (\Lambda _\pi ),m_\eta ) ,\nonumber \\ C_{2,\eta '}^{(*)}= & {} C_2(m_\pi ,m_\pi (\Lambda _\pi ),m_{\eta '}) ,\nonumber \\ C_{2,K}^{(*)}= & {} C_2(m_K,m_K(\Lambda _K),m_K) . \end{aligned}$$
(10)

However, the \(C_1\) terms have been disappearing due to \(p_\rho ^\mu \cdot \epsilon =0\) in the amplitudes.

For the three-body decays \(D_{s}^{+}\rightarrow \rho ^{0(+)}(a^{+(0)}_{0}\rightarrow ) \eta \pi ^{+(0)}, K^+{{\bar{K}}}^0(K^+ K^-)\), we present [7]

$$\begin{aligned}&\Gamma (D_{s}^{+}\rightarrow \rho a_0,a_0\rightarrow \eta \pi (KK))\nonumber \\&\quad =\int \int \frac{1}{(2\pi )^3}\frac{|\mathbf {p}_{2}|^{2}}{32m_{D_{s}} m_{\rho }^2}|H|^2 \frac{g_{\eta (K)}^{2}}{\mid D_{a_{0}}(s)\mid ^{2}}ds dt , \end{aligned}$$
(11)

with \(s=(p_{\eta (K)}+p_{\pi (K)})^{2}\) and \(t=(p_{\rho }+p_{\pi (K)})^{2}\), such that the theoretical results can be compared to the data in Eq. (1). In the above equation, \(1/D_{a_{0}}(s)\) presents the propagator for \(a_0\), and we define [26]

$$\begin{aligned}&D_{a_{0}}(s)=s-m_{a_0}^{2}-\sum \limits _{\alpha \beta } { [\text {Re}\Pi ^{\alpha \beta }_{a_0}(m^2_{a_0})-\Pi ^{\alpha \beta }_{a_0}(s) ]} , \end{aligned}$$
(12)

where

$$\begin{aligned} \Pi ^{\alpha \beta }_{a_0}(x)= & {} \frac{G^{2}_{\alpha \beta }}{16\pi } \bigg \{\frac{m_{\alpha \beta }^+ m_{\alpha \beta }^-}{\pi x} \log \bigg [\frac{m_\beta }{m_\alpha }\bigg ]-\theta [x-(m_{\alpha \beta }^+)^2]\nonumber \\&\times \rho _{\alpha \beta }\bigg (i+\frac{1}{\pi } \log \bigg [\frac{\sqrt{x-(m_{\alpha \beta }^+)^{2}}+\sqrt{x-(m_{\alpha \beta }^-)^{2}}}{\sqrt{x-(m_{\alpha \beta }^-)^{2}}-\sqrt{x-(m_{\alpha \beta }^+)^{2}}}\bigg ]\bigg ) \nonumber \\&-\rho _{\alpha \beta }\bigg (1-\frac{2}{\pi } \arctan \bigg [\frac{\sqrt{-x+(m_{\alpha \beta }^+)^{2}}}{\sqrt{x-(m_{\alpha \beta }^-)^{2}}}\bigg ]\bigg )\nonumber \\&\quad \times (\theta [x-(m_{\alpha \beta }^-)^{2}]-\theta [x-(m_{\alpha \beta }^+)^{2}])\nonumber \\&+\rho _{\alpha \beta }\frac{1}{\pi } \log \bigg [\frac{\sqrt{(m_{\alpha \beta }^+)^{2}-x}+\sqrt{(m_{\alpha \beta }^-)^2-x}}{\sqrt{(m_{\alpha \beta }^-)^{2}-x}-\sqrt{(m_{\alpha \beta }^+)^2-x}}\bigg ]\nonumber \\&\quad \times \theta [(m_{\alpha \beta }^-)^{2}-x]\bigg \} ,\nonumber \\ \rho _{\alpha \beta }\equiv & {} \left| \sqrt{x-(m_{\alpha \beta }^+)^{2}}\sqrt{x-(m_{\alpha \beta }^-)^{2}}\right| /x , \end{aligned}$$
(13)

with \(G_{\alpha \beta }=(g_{\eta ^{(\prime )}},g_K)\) for \(\alpha \beta ={(\eta ^{(\prime )}\pi ,K {{\bar{K}}})}\) and \(m_{\alpha \beta }^\pm =m_\alpha \pm m_\beta \).

3 Numerical results

In the numerical analysis, we adopt \(V_{cs}=V_{ud}=1-\lambda ^2/2\) with \(\lambda =0.22453\pm 0.00044\) in the Wolfenstein parameterization [7], along with \(m_{a_0^0}=0.987\) GeV [27, 28]. The topological parameters (TCA) have been extracted as [6]

$$\begin{aligned} (|T|, |C|, |A|)= & {} (0.363\pm 0.001,0.323\pm 0.030,0.064\nonumber \\&\pm 0.004)~\text{ GeV}^{3}~ ,\nonumber \\ (\delta _C,\delta _A)= & {} (-151.3\pm 0.3,23.0^{+\;\;7.0}_{-10.0})^{\circ } , \end{aligned}$$
(14)

where \(\delta _{C,A}\) are the relative strong phases. According to the extraction, we obtain \({{\mathcal {B}}}(D_s^+\rightarrow \pi ^{+}\eta ,\pi ^+\eta ^{\prime },K^{+}{\bar{K}}^0)= (1.8\pm 0.2,4.2\pm 0.5,3.1\pm 0.5)\times 10^{-2}\), consistent with the experimental values of \((1.68\pm 0.10,3.94\pm 0.25,2.95\pm 0.14)\times 10^{-2}\), respectively [7]. For \(V,S\rightarrow PP\), the strong coupling constants read [7, 27, 28]

$$\begin{aligned} g_{\rho }= & {} 6.0 , (g_\eta ,g_{\eta '},g_K)=(2.87\pm 0.09,-2.52\nonumber \\&\quad \pm 0.08,2.94\pm 0.13)~\text {GeV} . \end{aligned}$$
(15)

Empirically, \(\Lambda _M\) of \({{\mathcal {O}}}(1.0~\text {GeV})\) is commonly used to explain the data [29,30,31]; besides, it is obtained that \(\Lambda _K-\Lambda _\pi =m_K-m_\pi \) [32]. Therefore, we are allowed to use \((\Lambda _\pi ,\Lambda _K)=(1.25\pm 0.25,1.60\pm 0.25)\) GeV, which result in

$$\begin{aligned} C_{\eta }= & {} [(0.57\pm 0.08)+i(0.17\pm 0.09)]~\text {GeV}^{-2} ,\nonumber \\ C_{\eta ^{\prime }}= & {} [(0.34\pm 0.03)-i(0.27\pm 0.07)]~\text {GeV}^{-2} ,\nonumber \\ C_{K}= & {} [(0.85\pm 0.03)-i(0.45\pm 0.08)]~\text {GeV}^{-2} , \end{aligned}$$
(16)

with \(p_{a_0}^2=m_{a_0}^2\). Subsequently, we predict

$$\begin{aligned}&{{\mathcal {B}}}(D_{s}^{+}\rightarrow \rho ^{0(+)}a^{+(0)}_{0}) =(3.0\pm 0.3\pm 1.0)\times 10^{-3} ,\nonumber \\&{{\mathcal {B}}}(D_{s}^{+}\rightarrow \omega a^{+}_{0})=0 , \end{aligned}$$
(17)

where the first error in \(\mathcal{B}(D_{s}^{+}\rightarrow \rho ^{0(+)}a^{+(0)}_{0})\) takes into account the uncertainties from \(\Lambda _{\pi }\) and \(\Lambda _{K}\), and the second one combines those from \(V_{cs}^*\), \(V_{ud}\), (TCA), and the strong coupling constants. For the resonant three-body decays, we obtain

$$\begin{aligned}&{{\mathcal {B}}}(D_{s}^{+}\rightarrow \rho ^{0(+)}(a^{+(0)}_{0}\rightarrow )\eta \pi ^{+(0)})\nonumber \\&\quad = (1.6^{+0.2}_{-0.3}\pm 0.6)\times 10^{-3} ,\nonumber \\&{{\mathcal {B}}}(D_{s}^{+}\rightarrow \rho ^{+}(a^{0}_{0}\rightarrow )K^+K^-,K^0{{\bar{K}}}^0)\nonumber \\&\quad = (0.9^{+0.1}_{-0.1}\pm 0.4,0.7^{+0.1}_{-0.1}\pm 0.3)\times 10^{-4} ,\nonumber \\&{{\mathcal {B}}}(D_{s}^{+}\rightarrow \rho ^{0}(a^{+}_{0}\rightarrow )K^+{\bar{K}}^0)\nonumber \\&\quad =(1.5^{+0.2}_{-0.3}\pm 0.6)\times 10^{-4} , \end{aligned}$$
(18)

where the sources of the two errors are the same as those in Eq. (17).

4 Discussions and conclusions

In the triangle loop, when the momentum flow approaches the mass shell for one of the three propagators, the integration with \(i\epsilon \simeq im\Gamma \) gives rise to the imaginary parts in Eq. (16), where \(\Gamma \) is a very tiny decay width for \(\pi \), \(\eta ^{(\prime )}\) or K. The off-shell integrations are responsible for the real parts in Eq. (16), for which we take \({{\mathcal {M}}}_a\) as our description. In principle, the integration allows a momentum flow from \(-\infty \) to \(+\infty \). However, when the exchange particle proceeds with \(q_2^2\) around several GeV\(^2\), instead of the infinity, the integration with \(q_2^2\sim \pm \infty \) causes an overestimation [21]. We have accordingly introduced the form factor \(F_\pi \) in Eq. (6) to cut off the contribution from \(q_2^2\sim \pm \infty \). While \(q_1^2\), \(q_2^2\) and \(q_3^2\) have been associated in the loop, \(F_\pi \) also works to cut off the contributions from the propagators of the rescattering particles \(\pi ^+\) and \(\eta ^{(\prime )}\). Note that the single cutoff form factor has also been commonly used elsewhere [33,34,35].Footnote 1

The smallness of the WA \(D_s^+\rightarrow MM\) decay can be traced back to its amplitude, given by [37,38,39]

$$\begin{aligned}&{{\mathcal {M}}}_{WA}(D_s^+\rightarrow MM) \propto f_{D_s} q^\mu \langle MM|{{\bar{u}}}\gamma _\mu (1-\gamma _5) d|0\rangle \nonumber \\&\quad \simeq (m_u+m_d)\langle MM|{{\bar{u}}}\gamma _5 d|0\rangle , \end{aligned}$$
(19)

where \(q^\mu \langle MM|u\gamma _\mu d|0\rangle =0\) corresponds to the conservation of the vector current (CVC); most importantly, \(m_{u(d)}\simeq 0\) causes the chiral suppression of the WA \(D_s^+\rightarrow MM\) decay. According to the data, \({{\mathcal {B}}}(D_s^+\rightarrow \pi ^+\rho ^0)=(1.9\pm 1.2)\times 10^{-4}\) indicates that \(\mathcal{B}_{WA}(D_s^+\rightarrow MM)\) should be around \(10^{-4}\). In addition, \({{\mathcal {B}}}(D_s^+\rightarrow \pi ^+\pi ^0)<3.4\times 10^{-4}\) suggests that the G-parity violation suppresses the WA process even more. Therefore, since \(D_s^+\rightarrow \rho a_0,\rho \pi \) are both the G-parity conserved processes, it is reasonable to present that \({{\mathcal {B}}}_{WA}(D_s^+\rightarrow \rho a_0) \simeq {{\mathcal {B}}}_{WA}(D_s^+\rightarrow \rho \pi )\sim 10^{-4}\). As a theoretical support, we present

$$\begin{aligned}&{{\mathcal {B}}}_{WA}(D_s^+(c{{\bar{s}}})\rightarrow \rho a_0)\nonumber \\&\quad =R_f\frac{(V_{cs}^* V_{ud}f_{D_s})^2\tau _{D_s}}{(V_{cb}V_{ud}^* f_{B_c})^2\tau _{B_c}} {{\mathcal {B}}}_{WA}(B_c^+(c{{\bar{b}}})\rightarrow \rho a_0) \sim 10^{-4} , \end{aligned}$$
(20)

in agreement with our estimation, where \(R_f=3.1\times 10^{-2}\) is mostly from the phase space factors, and \({{\mathcal {B}}}_{WA}(B_c^+\rightarrow \rho a_0)\simeq 1.0\times 10^{-5}\) is adopted from Ref. [40].

Disregarding the WA contributions, we predict \(\mathcal{B}(D_{s}^{+}\rightarrow \rho ^{0(+)}a^{+(0)}_{0})=3.0\times 10^{-3}\) in Eq. (17). It is found that the \(\eta \pi ,\eta '\pi ,K^+\bar{K}^0\) rescatterings and their interferences give 6%, 7%, 30% and 59% of \({{\mathcal {B}}}(D_s^+\rightarrow a_0^{+(0)}\rho ^{0(+)})\), respectively. By contrast, the \(\rho \eta ^{(\prime )}\) rescatterings from \(D_s^+\rightarrow \rho \eta ^{(\prime )}\) dominantly contribute to \(D_s^+\rightarrow a_0^{+(0)}\pi ^{0(+)}\), instead of \(D_s^+\rightarrow K^* K\) [9]. Since \(D_{s}^{+}\rightarrow \omega a^{+}_{0}\) has no rescattering effects; besides, the WA contribution is suppressed by the G-parity violation, we anticipate that \(\mathcal{B}(D_{s}^{+}\rightarrow \omega a^{+}_{0}) \simeq \mathcal{B}(D_s^+\rightarrow \pi ^+\pi ^0)<3.4\times 10^{-4}\) [7].

In Eq. (18), \(\mathcal{B}(D_{s}^{+}\rightarrow \rho ^{0}(a^{+}_{0}\rightarrow )\eta \pi ^{+}) =(1.6^{+0.2}_{-0.3}\pm 0.6)\times 10^{-3}\) is able to explain the data [see Eq. (1)], demonstrating the sufficient long-distance annihilation contribution. It is confusing that \(\mathcal{B}(D_s^+\rightarrow a_0^0 \rho ^+,a_0^0\rightarrow K^+ K^-)\) is 10 times smaller than \({{\mathcal {B}}}_0\) in Eq. (1). For clarification, we take the approximate form of the resonant branching fraction: \(\mathcal{B}(D_s^+\rightarrow SV,S\rightarrow PP)\simeq {{\mathcal {B}}}(D_s^+\rightarrow SV){{\mathcal {B}}}(S\rightarrow PP)\), together with the isospin relation: \(\mathcal{B}(D_{s}^{+}\rightarrow \rho ^0 a^+_0)\simeq {{\mathcal {B}}}(D_{s}^{+}\rightarrow \rho ^+ a^0_{0})\), such that \({{\mathcal {B}}}_0/{{\mathcal {B}}}_+\) is reduced as \(\mathcal{B}(a_0^0\rightarrow K^+ K^-)/{{\mathcal {B}}}(a_0^+\rightarrow \eta \pi ^+)\simeq 1/3\), disagreeing with \({{\mathcal {B}}}(a_0^0\rightarrow K^+ K^-)/{{\mathcal {B}}}(a_0^+\rightarrow \eta \pi ^+)=1/10\) from the experimental extraction [8]. Therefore, we conclude that there exists a possible contradiction between the observations in Eq. (1).

In our reasoning, the contradiction might be caused by \(D_s^+\rightarrow \rho ^+ f_0,f_0\rightarrow K^+ K^-\) with \(f_0\equiv f_0(980)\), which can be mistaken as \(D_s^+\rightarrow \rho ^+ a_0^0,a_0^0\rightarrow K^+ K^-\) [41]. First, since \(D_s^+\rightarrow \rho ^+ f_0\) is an external W-boson emission process, its branching fraction can be of order \(10^{-3}\). Second, \(a_0\) and \(f_0\) are both scalar mesons, and have nearly the same masses and overlapped decay widths. As a result, it is possible that one cannot distinguish between the resonant signals of \(D_s^+\rightarrow \rho ^+(f_0,a_0), (f_0,a_0)\rightarrow K^+ K^-\) in the \(K^+ K^-\) invariant mass spectrum. For a careful examination, we suggest a measurement of \({{\mathcal {B}}}(D_{s}^{+}\rightarrow \rho ^{0}(a^{+}_{0}\rightarrow )K^+{\bar{K}}^0)/ \mathcal{B}(D_{s}^{+}\rightarrow \rho ^{+}(a^{0}_{0}\rightarrow )K^+ K^-)\), which will be observed around 2 if there exists no resonant \(f_0\rightarrow K^+ K^-\) decay to be involved in \({{\mathcal {B}}}(D_{s}^{+}\rightarrow \rho ^+ K^+ K^-)\).

In summary, we have studied \(D_{s}^{+}\rightarrow \rho ^{0(+)}a^{+(0)}_{0}\), \(D_{s}^{+}\rightarrow \omega a_0^0\), and the resonant \(D_{s}^{+}\rightarrow \rho a_0\), \(a_{0}\rightarrow \eta \pi (KK)\) decays. In the final state interaction, where \(D_s^+\rightarrow \eta ^{(\prime )}\pi ^+ (K^+{{\bar{K}}}^0)\) is followed by the \(\eta ^{(\prime )}\pi ^+\) (\(K^+{{\bar{K}}}^0\)) to \(\rho a_0\) rescattering, we have predicted \(\mathcal{B}(D_{s}^{+}\rightarrow \rho ^{0(+)}a^{+(0)}_{0})=(3.0\pm 0.3\pm 1.0)\times 10^{-3}\). Because of the non-contribution of the rescattering effects and the suppressed short-distance W annihilation, it has been expected that \({{\mathcal {B}}}(D_{s}^{+}\rightarrow \omega a^{+}_{0}) \simeq {{\mathcal {B}}}(D_s^+\rightarrow \pi ^+\pi ^0)<3.4\times 10^{-4}\). For the resonant three-body decay, \(\mathcal{B}(D_{s}^{+}\rightarrow \rho ^{0}(a^{+}_{0}\rightarrow )\eta \pi ^{+}) =(1.6^{+0.2}_{-0.3}\pm 0.6)\times 10^{-3}\) has been shown to agree with the data. We have presented \(\mathcal{B}(D_{s}^{+}\rightarrow \rho ^{+}(a^{0}_{0}\rightarrow )K^+K^-)\) 10 times smaller than the observation, indicating that a more careful examination is needed.