Chapter 0 Hadronic molecules and multiquark states

Christoph Hanhart Forschungszentrum Jülich, Institute for Advanced Simulation, D-52425 Jülich, Germany

Chapter Article tagline: update of previous edition, reprint.

Abstract

[Abstract] In this section different theoretical approaches towards multi-quark states are introduced, namely hadrocharmonia, compact tetraquarks and hadronic molecules. The predictions derived from either of them are contrasted with current and possible future observations. The focus is on doubly heavy systems, but singly heavy and light systems are mentioned briefly as well.

Figure 1: Visualisation of the substructure of possible tetraquark configurations in the doubly heavy sector, with $Q$ ($\bar{Q}$) and $q$ ($\bar{q}$) denoting heavy and light (anti-)quarks, respectively: (a) hadroquarkonium, (b) compact tetraquark, (c) hadronic molecule. While the first two are typically compact with their size dictated by the confinement radius, the last one can be very large if the state is located close to the pertinent two-hadron threshold.

(a)                                 (b)                                 (c)

Since the early days of the quark model, multiquark states are proposed to exist—they are already mentioned in the famous works by Gell-Mann [1] and Zweig [2]. The first calculation proposing a multiquark structure for concrete states was performed by Jaffe and Johnson within the MIT bag model: The nonet of light scalar mesons, nowadays called $f_{0}(500)$, $f_{0}(980)$, $a_{0}(980)$ and $K^{*}_{0}(700)$, were proposed to be tetraquarks ($\bar{q}\bar{q}qq$) [3]. Only one year later Voloshin and Okun [4] argued for the existence of deuteron like states in doubly heavy systems. The idea was picked up later by Törnqvist [5]. In the previous section strong experimental evidence was presented for the existence of states beyond the most simple structures allowed by the rules for the formation of hadrons within QCD, especially in the doubly heavy sector (for reviews, putting emphasis on different aspects, see Refs. [6, 7, 8, 9, 10, 11, 12, 13, 14])—this is why we start the discussion for this class of states that we use to introduce the various structure assumptions currently discussed in the literature. We will come back to candidates for multiquark states in other sectors towards the end of this section. In the doubly heavy sector there were states found that decay into a heavy quarkonium state together with a light hadron. Examples are $T_{\bar{b}b}(10610)$ and $T_{\bar{b}b}(10650)$, also known as $Z_{b}$ states, observed in the $\Upsilon(nS)\pi$, $n=1,2,3$, and $h_{b}(mP)\pi$, $m=1,2$, final states and $P_{c\bar{c}}(4312)$, $P_{c\bar{c}}(4440)$ and $P_{c\bar{c}}(4457)$ observed in $J/\psi p$ final states. Since the production of the heavy quarkonium in the course of the decay is heavily suppressed within QCD due to the OZI rule, the $\bar{Q}Q$ pair must have preexisted in the wave functions of the observed states. Since they on the other hand carry charge (or, equivalently, non-vanishing isospin), they are identified as tetraquarks and pentaquarks, respectively.

In addition to the explicit multiquark states mentioned in the previous paragraph there are also states that qualify for multiquark states not because of their quantum numbers, but because of their unusual properties, difficult if not impossible to accommodate within the most simple realisations of the quark model or variants thereof. Prominent examples of this class are the $\psi(4230)$, also known as $Y(4230)$, a vector state that decays, e.g., into $D\bar{D}^{*}\pi$ and $J/\psi\pi\pi$ but not into $D^{(*)}\bar{D}^{(*)}$ as is expected for a $\bar{c}c$ state, and the $\chi_{c1}(3872)$, also known as $X(3872)$, which decays by far dominantly into $D^{0}\bar{D}^{*\,0}$, although its mass basically coincides with the threshold of this channel, while the decays into $J/\psi\pi\pi$ and $J/\psi 3\pi$ appear to be heavily suppressed.

In this section the different proposals put forward for the structure those states, sketched for tetraquarks in Fig. 1, are reviewed. The presentation focusses on what imprint in the particle spectrum and observables the different structures would leave, if they were to be the only or the by far dominant component in a given state. This is also what is mostly discussed in the literature as of today. However, in principle some mixing between the different structures is possible as well. We come back to this interesting question in section 8. The different structures shown in Fig. 1 differ by the assumed sub-structures within the hadron. Those are either color neutral (for hadro-quarkonia or hadronic molecules) or carrying a color charge in form of (anti-)diquarks.

The link between QCD and the different assumed structures of the multiquark states is provided by the approximate symmetries of QCD and their breaking, most notably SU(2) flavor (also known as isospin symmetry), embedded in the larger SU(3) flavor symmetry and heavy quark spin symmetry.

In addition, also chiral symmetry plays a crucial role for some systems. SU(2) (SU(3)) flavor symmetry were an exact symmetry of QCD, if the charges and masses of up and down (up, down and strange) quarks were equal, since the QCD interaction is flavor blind (up to some quark type dependence entering through the scale dependence of the QCD coupling constant). Since the up and down quark mass difference is much smaller than any hadronic scale, isospin symmetry is typically realised with an accuracy of better than a few percent. As the strange quark is much heavier, SU(3) flavor symmetry is realised only with some 30% accuracy, however, SU(3) breaking mass differences within multiplets of compact states are typically (largely) explained by the mass added in by the strange quark. For hadronic molecules the situation is more complicated as is discussed below.

Heavy quark spin symmetry is a consequence of spin dependent interactions scaling as $q_{\rm typ}/M_{Q}$, where $M_{Q}$ denotes the mass of the heavy quark and the typical momentum inside a hadron may be estimated by the non-perturbative QCD scale, $q_{\rm typ}\sim\Lambda_{\rm QCD}$. In the strict heavy quark limit ($M_{Q}\to\infty$), spin multiplets appear formed from the heavy quark spin being coupled differently to the same light quark cloud. In other words, in this limit a light quark cloud of some $j_{\ell}$ (this quantum number should also contain some possible orbital angular momentum among heavy quarks in the system) coupled to some heavy quark spin $S_{h}$ will generate a whole set of states with total angular momentum

$$|j_{\ell}-S_{h}|\leq J\leq j_{\ell}+S_{h}\ .$$

In systems with more than one heavy quark the multiplets get even larger, since the properties of the emerging hadrons in leading order of the expansion do not depend on the value of the total spin $S_{h}$ the spins of the different heavy quarks in the system get coupled to. Since the symmetry is exact in the heavy quark limit it is realised in the hadron spectrum regardless the assumed structure of it. However, the symmetry violations turn out to be structure dependent [15] are expected to become an important diagnostic to deduce the structure of exotic hadrons from the spectra.

Theoretical approaches to multiquark states range from phenomenological models, either on the quark level or of meson exchange type, over effective field theories, again on the quark level or employing chiral perturbation theory on the hadron level, to lattice QCD, each of which are described in other sections of this encyclopedia. In what follows some account will be given on their application to multiquark states.

1 Hadroquarkonia

For tetraquarks the assumed structure of a hadroquarkonium is shown in Fig. 1(a). The underlying idea is that the multiquark consists of a compact, color neutral $Q\bar{Q}$ core, like the $J/\psi$, the $\psi(2S)$ or the $\eta_{c}$ surrounded by a (typically excited) light quark cloud [16]. Since a $Q\bar{Q}$ state does not contain light quarks, the interaction of the cloud with the compact doubly heavy core is suppressed (e.g. at leading order chiral perturbation theory the interaction of pions with quarkonia vanishes), however, the polarisabilities might still be sufficiently large to allow for some binding of a light quark cloud to e.g. $J/\psi$ or excitations thereof. The same mechanism might also provide sufficient binding, to generate states of pairs of $J/\psi$s [17]. We come back to this scenario in Sec. 5.

The proposal for the existence of a hadrocharmonium structure was triggered by the observation of $\psi(4230)$ also known as $Y(4230)$ (at the time called $Y(4260)$), since this state shows up as a clear peak in the $J/\psi\pi\pi$ spectrum while at the same time being absent in the spectra for $D\bar{D}$, $D\bar{D}^{*}$ and $D^{*}\bar{D}^{*}$. Such a pattern emerges naturally within the hadrocharmonium picture, since the observed decay is merely a fall apart mode of the building blocks, while a transition to the open charm channels requires a break up of the compact quarkonium as well as some re-arrangement of the light quark cloud. Note that recently the $Y(4230)$ was also observed as a clear peak in the reaction $e^{+}e^{-}\to\bar{D}\pi D^{*}$ with an even higher rate than in $J/\psi\pi\pi$ [18, 19], questioning somewhat the logic put forward above [20].

The $Y(4230)$ is also observed in the $h_{c}\pi\pi$ channel. In contrast to the $J/\psi$ that has a spin 1, the $h_{c}$ has spin 0. Thus, if the $Y(4230)$ were to contain a pure $J/\psi$ core, heavy quark spin symmetry would prevent it from decaying into a final state with a spin 0 quarkonium. To overcome this discrepancy with experiment, in Ref. [21] it was suggested that the $Y(4230)$ is not a pure state but, together with the next heavier state, the $\psi(4360)$ also known as $Y(4360)$, emerges from a spin symmetry violating mixing of two states with $h_{c}$ and $\psi(2S)$ cores, respectively. Here it is assumed that the suppression of the spin symmetry violation which is shown to appear already at order $\Lambda_{\rm QCD}/m_{c}$, is overcome by the close proximity of the mixing states leading to a small energy denominator in the mixing amplitude. More concretely, starting from the unmixed basis

$$\Psi_{3}=(1^{--})_{c\bar{c}}\otimes(0^{++})_{q\bar{q}}\qquad\mbox{and}\qquad\Psi_{1}=(1^{+-})_{c\bar{c}}\otimes(0^{-+})_{q\bar{q}}\ ,$$ (1)

where the heavy cores are assumed to be $\psi(2S)$, with a mass of 3686 MeV, and $h_{c}(1P)$, with a mass of 3525 MeV, one gets for the physical states

$$Y(4230)=\cos(\theta)\Psi_{3}-\sin(\theta)\Psi_{1}\qquad\mbox{and}\qquad Y(4360)=\sin(\theta)\Psi_{3}+\cos(\theta)\Psi_{1}\ .$$ (2)

A fit to data revealed a mixing angle of the order of 40 degrees accompanied by near degenerate unmixed states with masses of approximately 4.30 and 4.32 GeV for $\Psi_{3}$ and $\Psi_{1}$, respectively. The apparent close proximity of the mixed states must emerge somewhat accidental from the interplay of the core states that show a mass difference of the order of 160 MeV with the light quark clouds.

Figure 2: Illustration of the implications of employing spin symmetry to predict the spin partners of the $Y(4230)$ and $Y(4360)$ in the hadrocharmonium scenario. Solid lines show masses of the physical states (input or predicted), dashed line the deduced masses of the unmixed basis states. The $1^{--}$ states are used as input — the masses of the predicted spin partner states are approximate only. Figure adapted from Ref. [15].

What testable predictions emerge from this scenario? If spin symmetry were exact and a given exotic were a hadroquarkonium, it would imply that replacing the core of such a state by its spin partner(s) while keeping the light quark cloud the same, must lead to another hadroquarkonium state, whose mass can be estimated from the mass differences between the different seed states. While one can expect some spin symmetry violation in systems with charm, predictions derived from spin symmetry, allowing for the above-mentioned mixing, should capture the relevant patterns emerging from the assumed structure. This idea was exploited quantitatively in Ref. [15]—the spin partner states here are found by the replacements

$$\psi(2S)\to\eta_{c}(2S)\qquad\mbox{and}\qquad h_{c}(1P)\to\{\chi_{c0}(1P),\ \chi_{c1}(1P),\ \chi_{c2}(1P)\}\ .$$ (3)

The emerging spectrum is shown in Fig. 2. As a solid prediction of the hadrocharmonium scenario, a relatively light exotic $\eta_{c}$ state is predicted at 4.1 GeV that should not decay to $D\bar{D}^{*}$ but instead to $\eta_{c}\pi\pi$. Moreover, at about 4.3 GeV there should be a spin exotic state with quantum numbers $1^{-+}$, which cannot be generated for $\bar{c}c$ states, that is near degenerate with another $\eta_{c}$ state. This spectrum nicely shows how using spin symmetry allows for testable predictions for assumed underlying structures for the exotic states. Similar arguments can be applied to decay patterns. For example, in [22] it is argued that, if $Z_{c}(4100)$ and $Z_{c}(4200)$, also known as $T_{c\bar{c}1}(4100)^{+}$ and $T_{c\bar{c}1}(4200)^{+}$, respectively, were hadrocharmonia with seeds $\eta_{c}(2S)$ and $\psi(2S)$, respectively, their partial widths to final states containing the core state and a pion should be equal up to spin-symmetry violating corrections.

The building blocks of hadrocharmonia are color neutral by construction, since the core states are conventional charmonia. Thus, one might be tempted to put them into one class with hadronic molecules to be discussed in Sec. 3. However, while hadronic molecules typically have masses close to the thresholds of the channels that form the molecule, which can result in hadrons of unusually large size, hadroquarkonia can be in mass quite far above the most pertinent threshold and accordingly show typical hadronic sizes. For example, the $Y(4230)$ mentioned above is more than 600 MeV heavier than the sum of the masses of the $J/\psi$ and the lightest scalar-isoscalar resonance, the $f_{0}(500)$. For the $Y(4230)$ the hadrocharmonium and hadronic molecule picture are contrasted in Ref. [20]. The only case discussed so far in the literature that could be viewed as both a hadronic molecule and a hadrocharmonium is $\psi(4660)$ also known as $Y(4660)$, which is proposed to have a substructure made of a $\psi(2S)$ and an $f_{0}(980)$ [23, 16]. A prediction that will allow for a test of this proposal is that, if the assumed structure is correct, there needs to be a pseudoscalar bound state formed of $\eta_{c}(2S)$ and $f_{0}(980)$, located at around 4616 MeV (lighter than the $Y(4660)$ by the $\psi(2S)$-$\eta_{c}(2S)$ mass difference) with a partial width into $\eta_{c}(2S)\pi\pi$ of $60\pm 30$ MeV [24].

Clearly, for doubly heavy states of the type $QQ\bar{q}\bar{q}$ like the $T_{cc}(3875)^{+}$ discovered at LHCb, one may expect in the spirit of this section a compact $QQ$ core surrounded by a light $\bar{q}\bar{q}$ cloud. Thus, here the building blocks are (anti-)diquarks and thus qualify at the same time as compact tetraquarks. This kind of states is discussed in some detail at the end of the next section.

2 Compact Tetraquarks

The crucial building blocks of compact tetra- and pentaquarks are (anti-)diquarks—for a review about diquark properties we refer to Ref. [25]. Since quarks (antiquarks) live in the color $[3]$ ($[\bar{3}]$) representation, a quark (antiquark) pair lives either in the color $[\bar{3}]$ or the color $[6]$ ($[3]$ or $[\bar{6}]$) representation. In many works as well as here, only the diquarks in the color $[\bar{3}]$ representation are being kept, motivated, e.g., by the observation that at least for the one gluon exchange only in this quark-quark channel the interaction is attractive. Thus (anti-)diquarks carry a color charge and are therefore subject to confinement—the hadrons with these as building blocks are necessarily compact with their size given by the confinement radius. Model calculations reveal that the scalar (spin 0) diquarks are lighter than the axial-vector diquarks (spin 1)—this is why they were dubbed good and bad diquarks, respectively. With this being said, one finds for the basis states forming doubly heavy tetraquarks [26] two states with $S_{h}^{C}=0^{+}$,

$$|0^{+}\rangle=|0_{cq},0_{\bar{c}\bar{q}};S_{h}=0\rangle\ ,\ |0^{+^{\prime}}\rangle=|1_{cq},1_{\bar{c}\bar{q}};S_{h}=0\rangle\ ,$$ (4)

three states with $S_{h}=1$

$$|A\rangle=|1_{cq},0_{\bar{c}\bar{q}};S_{h}=1\rangle\ ,\ |B\rangle=|0_{cq},1_{\bar{c}\bar{q}};S_{h}=1\rangle\ ,\ |C\rangle=|1_{cq},1_{\bar{c}\bar{q}};S_{h}=1\rangle\ ,$$ (5)

that can be combined into states with defined charge parity $C$ as

$$|1^{+}\rangle=\frac{1}{\sqrt{2}}(|A\rangle+|B\rangle)\ ,\ |1^{-}\rangle=\frac{1}{\sqrt{2}}(|A\rangle-|B\rangle)\ ,\ |1^{-}\rangle=|C\rangle\ ,\$$ (6)

and finally one state with $S_{h}^{C}=2^{+}$, namely

$$|2^{+}\rangle=|1_{cq},1_{\bar{c}\bar{q}};S_{h}=2\rangle\ .$$ (7)

If the (anti-)diquark structures can be treated as well defined building blocks of the emerging hadrons, to estimate the spectrum of compact tetraquarks one can straightforwardly write down the pertinent interactions within the given hadron in analogy to what is known from atomic physics as

$$H=2m_{[cq]}+2(\kappa_{cq})_{\bar{3}}[(\vec{S}_{c}\cdot\vec{S}_{q})+(\vec{S}_{\bar{c}}\cdot\vec{S}_{\bar{q}})]+2(\kappa_{c\bar{q}})[(\vec{S}_{c}\cdot\vec{S}_{\bar{q}^{\prime}})+(\vec{S}_{\bar{c}}\cdot\vec{S}_{q})]+2\kappa_{\bar{q}q}(\vec{S}_{q}\cdot\vec{S}_{\bar{q}^{\prime}})+2\kappa_{\bar{c}c}(\vec{S}_{c}\cdot\vec{S}_{\bar{c}})+\frac{B_{\cal Q}}{2}\vec{L}^{2}+2a_{Y}\vec{L}\cdot\vec{S}+b_{Y}S_{12}\ ,$$ (8)

where the first term captures the diquark masses and the next four account for the spin-spin interactions between the different quarks in the hadron. The following two terms need to be added to describe states with inter-diquark angular momenta larger than zero [27], where $\vec{S}=\vec{S}_{\mathcal{Q}}+\vec{S}_{\mathcal{\bar{Q}}}$ with $\vec{S}_{\mathcal{Q}}=\vec{S}_{c}+\vec{S}_{q}$ and analogously for $\vec{S}_{\mathcal{\bar{Q}}}$. Finally, the tensor-operator is defined via

$$S_{12}=3(\vec{S}_{\mathcal{Q}}\cdot\vec{n})(\vec{S}_{\mathcal{\bar{Q}}}\cdot\vec{n})-\vec{S}_{\mathcal{Q}}\cdot\vec{S}_{\mathcal{\bar{Q}}}\ ,$$

where $\vec{n}$ is either $\vec{r}/r$ or $\vec{q}/q$ for calculations in coordinate space or momentum space, respectively, with $\vec{r}$ the distance between the building blocks and $\vec{q}$ the momentum transfer. The matrix elements of the above operators can be evaluated with standard methods and the strength parameters need to be determined from experiment. The label on the first spin-spin term indicates that in the model sketched here the diquarks are considered in the color $[\bar{3}]$ representation only. Heavy quark spin symmetry suggests that $\kappa_{cq}/\kappa_{q\bar{q}}\sim M_{q}/M_{c}$ and $\kappa_{c\bar{c}}/\kappa_{q\bar{q}}\sim(M_{q}/M_{c})^{2}$.

We start with a discussion of $S$ wave states ($L=0$). The Hamiltonian of Eq. (8) is diagonal in the $J^{PC}=1^{++}$ and $2^{++}$ channels and Ref. [26] quotes

$$M(2^{++})=M(1^{++})+2[(\kappa_{cq})_{\bar{3}}+\kappa_{c\bar{q}}]=3952\ \mbox{MeV}\ ,$$ (9)

having identified the $1^{++}$ state with the $\chi_{c1}(3872)$ aka $X(3872)$ and estimated the various spin-spin interactions from the masses of regular charmonia. It is furthermore argued that the $2^{++}$ state can be identified with the tensor state observed at 3940 MeV. The given interaction predicts a $1^{+-}$ state at 3882 MeV which is nicely consistent with the mass of the isovector state $T_{c\bar{c}1}(3900)^{+}$ also known as $Z_{c}(3900)$. However, the corresponding spin partner state is predicted at 3754 MeV, significantly lighter than the $X(3872)$, while it is observed experimentally more than 100 MeV above, namely at 4020 MeV. To accommodate this, in Refs. [28, 29] it was proposed that contrary to the heavy quark spin symmetry scaling quoted above, all spin-spin interactions but the first should be negligible. A justification for this unexpected pattern was provided later in Ref. [30], where it was argued that the diquarks are separated by some potential well, no allowing the short ranged spin-spin interactions to operate between different diquarks—also the dynamical diquark picture proposed in Ref. [31] leads to a sizeable separation of diquark and anti-diquark within the exotic hadron. Then the strength parameter of the remaining spin-spin interaction is estimated from the mass difference of $Z_{c}(3900)$ and $Z_{c}(4020)$ aka $T_{c\bar{c}1}(3900)^{+}$ and $T_{c\bar{c}1}(4020)^{+}$, respectively, to be $(\kappa_{cq})_{\bar{3}}=67$ MeV [28], about three times larger than the value provided in Ref. [26], where it was extracted from the $\Sigma_{c}$-$\Lambda_{c}$ mass difference. In this way the $2^{++}$ and the second $1^{+-}$ state are predicted to have a mass quite close to the $D^{*}\bar{D}^{*}$ threshold (in line with the prediction in the molecular approach as detailed below).

Figure 3: Spectrum of negative parity tetraquarks predicted in Ref. [27]. Note that the masses of the $1^{--}$ states used as input are slightly different compared to those employed in the other sections, however, an adaption would not change the overall picture. Note that the quantum numbers $0^{--}$ and $1^{-+}$ are exotic—they cannot be reached from quark-antiquark states. Fig. from Ref. [11].

The building blocks for tetraquarks containing a heavy quark and its anti-quark with negative parity are still the ones provided in Eqs. (4), (6), and (7). The negative parity is then provided by an angular momentum between the diquark and the anti-diquark. For charge and flavor neutral states an odd angular momentum changes also the charge parity. Thus, one get e.g. five states with $J^{PC}=1^{--}$: Two by adding one unit of angular momentum to the two states provided in Eq. (4) or by coupling the total $S_{h}^{+}$ states of Eqs. (6) and (7) with an angular momentum of one to a total angular momentum of one. Finally the state given in Eq. (7) can be combined with $L=3$ to a total angular momentum of one. The $\vec{L}^{2}$-term pushes the vector state with $L=3$ so far up, that it does not need to be considered any further. One thus gets three additional parameters that can be fixed from the vector states. Alternatively one can include the parameter $(\kappa_{cq})_{\bar{3}}$ in the fit and see if a value consistent with that found for the $S$-wave states is found. This is the strategy followed in Ref. [27] and indeed, one of the fits is consistent with the existing data as well as the spin coupling term within 2$\sigma$ . The four vector states included in the fit were $Y(4230)$, $Y(4320)$, $Y(4390)$ and $Y(4660)$, see Fig. 3. Once the parameters are fixed, the masses of other exotics with different quantum numbers can be predicted, most strikingly the authors find two pseudoscalar states at about 4700 and 4270 MeV, respectively, and even three states with quantum numbers forbidden in the naive quark model, namely a $0^{--}$ state at a similar mass as the $Y(4230)$ and two near degenerate $1^{-+}$ states at about 4310 MeV.

As in the regular quark model, also in the compact tetraquark model radial excitations of the ground states should appear. For example a natural candidate for a radial excitation of $Z_{c}(3900)$ is the $Z_{c}(4430)$ [28] aka $T_{c\bar{c}1}(3900)^{+}$ and $T_{c\bar{c}1}(4430)^{+}$, respectively, a state difficult to understand in other approaches. Moreover, as in the standard quark model compact tetraquarks should fill complete SU(3) flavor multiplets, with the amount of SU(3) breaking being driven by the strange quark mass $M_{s}-M_{u}=120-150$ MeV . In particular, as long as no quark type dependence is included in the interaction hamiltonian, each isoscalar state is necessarily accompanied by a triplet of isovector states, with very similar mass—in analogy of the near degeneracy of the isovector $\rho$ and the isoscalar $\omega$ in the light quark sector. While this nicely explains the large isospin violation in the decays of $X(3872)$, it in effect leads to a larger number of states than observed in experiment so far, although there are indications that the SU(3) multiplets start to fill up [32]—note, however, that the experimental evidence as of today for the states that contain strangeness is rather weak. As a further refinement of the diquark–anti-diquark interaction in Ref. [29] an isospin dependent contribution is discussed.

Recently LHCb discovered a very intriguing state in the $D^{0}D^{0}\pi^{+}$ mass distribution, the $T_{cc}(3875)^{+}$ with a minimal quark content of $cc\bar{u}\bar{d}$. In the tetraquark model such states were discussed qualitatively in Ref. [33]—more quantitative investigations can be found in Refs. [34, 35]. Contrary to the heavy-light diquarks discussed above, for this state the constituents need to be heavy-heavy and light-light diquarks. Now the diquarks are subject to the Pauli principle. Therefore the charmed diquark needs to be in the spin one state. The light anti-diquark appears as the good combination, spin 0, in the anti-symmetric flavor $[3]$, while the bad, spin 1, anti-diquarks are in the symmetric flavor $[\bar{6}]$, since still only the totally antisymmetric color $[3]$ configuration is kept. Accordingly, the lightest tetraquark with the mentioned quark content needs to have the quantum numbers $J^{P}=1^{+}$ consistent with current experimental observations, although an amplitude analysis is still missing. The mass of the $T_{cc}(3875)^{+}$ basically coincides with the $D^{*+}D^{0}$ threshold. However, if indeed the $cc$-diquark forms an essential ingredient in this state one should expect the flavor partner of the $T_{cc}(3875)^{+}$ with $cc$ being replaced by $bb$ to quite deeply bound, simply since the heavier quark pair should sit closer together feeling the much stronger attraction provided by the one gluon exchange (the increase in strength driven by the $1/r$ scaling of the potential vastly overcomes the decrease in the strong coupling $\alpha_{s}$ which comes in since the pertinent mass scale rises). Already in the 80ties it was stressed that as soon as the ratio of heavy to light quark masses is well beyond 10, binding should occur for states of the kind discussed here [36, 37]. Note that these studies were performed within constituent quark models and accordingly we here talk about ratios of constituent masses, which for light quarks are of the order of 300 MeV not the elementary current quark masses of the order of a few MeV. Thus, based on those studies one should expect sizeable binding for the $bb\bar{q}\bar{q}$ systems only.

Figure 4: Comparison of the binding energies in MeV relative to the $BB^{*}$ threshold predicted for $bb\bar{u}\bar{d}$ tetraquarks with quantum numbers $I(J^{P})=0(1^{+})$ from different approaches. The columns show in order the spread in the predictions from different works, typically with much smaller individual uncertainties, for phenomenological models, QCD sum rules, lattice QCD employing static $QQ$ potentials and lattice QCD employing NRQCD for the heavy quarks.

By now there is a large number of studies for doubly heavy tetraquarks including phenomenological models [36, 37, 38, 34, 35, 39, 40, 41] lattice QCD either employing static potentials [42, 43, 44, 45] or NRQCD for the $b$-quarks [46, 47, 48, 49, 50, 51] as well as QCD sum rules [52, 53, 54] . They all agree that a $T_{bb}$ should be bound deeply in the $I=0$ and $J^{P}=1^{+}$ channel, although there is still a considerable spread in the predicted binding energies relative to the lowest open charm threshold ($BB^{*}$), which range from near zero to 500 MeV—see Fig. 4.

3 Hadronic Molecules

Atomic nuclei are well understood as bound systems of nucleons, protons and neutrons. The lightest non-trivial nucleus is the deuteron, a bound state of a single proton with a single neutron and a binding energy of only 2.2 MeV. A relatively strongly bound nucleus is the $\alpha$ particle, made of two protons and two neutrons — is has a binding energy of 7 MeV per nucleon which translates into a separation energy into two deuterons of 12 MeV. In addition there are bound systems of hyperons and nucleons, the so-called hypernuclei. Thus also there exists some SU(3) structure of hadron-hadron bound systems. The hypertriton, which may be viewed as a bound system of a $\Lambda$ baryon and a deuteron, has a $\Lambda$ separation energy of only 0.13 keV. We thus observe that nuclear binding energies span over two orders of magnitude, however, always stay well below 100 MeV. We may use this as some guidance for what to expect for hadronic molecules in general. Hadronic molecules are multi quark states whose substructure is given by color neutral hadrons. Thus they are the analogue of atomic nuclei, only that their constituents are different hadrons than ground state baryons.

Hadronic molecules are a special realisation of states one may want to classify as two-hadron states: The corresponding poles are generated through the hadron-hadron dynamics and not from e.g. gluon exchanges between quarks. The corresponding states may not only appear as near threshold bound or virtual states that qualify as hadronic molecules, but even wide resonances such as the $f_{0}(500)$ which has a mass of the same order of magnitude as its width can fall into this class (see the discussion in 7).

To provide some general understanding under what conditions the scattering of two particles of mass $m_{1}$ and $m_{2}$ that feel an attractive force develops a pole near threshold, we may study a zero range interaction—this is justified as soon as the de Broglie wave length connected to the bound system is significantly larger than the range of forces. Then the scattering amplitude at tree level in momentum space is given by a constant and we get for the loop expansion of the scattering amplitude in non-relativistic kinematics [8]

$$T_{\rm NR}=C\sum_{n=0}^{\infty}(C\Pi(E))^{n}=\frac{C}{1+C\Pi(E)}\qquad\mbox{with}\qquad\Pi(E)=-\int\frac{d^{3}q}{(2\pi)^{3}}\frac{F^{2}_{\Lambda}(q)}{E-q^{2}/(2\mu)+i\epsilon}=\frac{\mu}{2\pi}\left(\Lambda+i\sqrt{2\mu E+i\epsilon}\right)+{\mathcal{O}}(\Lambda^{-1})\ ,$$ (10)

where $F_{\Lambda}(q)$ is some regularisation function and $\Lambda$ the regularisation scale. For $F_{\Lambda}$ one may either chose a sharp cut off, $F_{\Lambda}(q)=\theta(\pi\Lambda/2-q)$, or a Gaussian cut-off or employ more sophisticated methods like the so-called power divergence subtraction [55] (for a detailed discussion see Ref. [56]). Here $\mu=m_{1}m_{2}/(m_{1}+m_{2})$ denotes the reduced mass of the external particles. The condition for the appearance of a near threshold pole in the amplitude may thus be expressed as

$$C^{-1}=-\Pi(-E_{B})=\frac{\mu}{2\pi}\left(-\Lambda+\gamma\right)+{\mathcal{O}}(\Lambda^{-1})\ ,$$ (11)

with the binding momentum $\gamma=\sqrt{2\mu E_{B}}$, where $E_{B}=m_{1}+m_{2}-m_{B}$ with $m_{B}$ for the mass of the studied bound state. Formally the scale $\Lambda$ is not an observable and needs to be absorbed into the coupling strength $C$. However, as soon as one would introduce more dynamics into the potential, $\Lambda$ would acquire a physical meaning like, e.g., the mass of the lightest exchange particle. Thus the appearance of a near threshold pole is controlled by some interplay of the binding potential, the masses of the external particles and the binding momentum. For $m_{1}\approx m_{2}$ we have $\mu\approx m_{2}/2$ and for $m_{1}\ll m_{2}$ we have $\mu\approx m_{1}$. Thus we deduce that it is more natural for doubly heavy systems to generate bound states than for singly heavy ones, since in the former case already a rather weak attraction ($|C|$ small) leads to a bound state.

The renormalised scattering amplitude, where $\Lambda$ is absorbed into $C$, takes the simple form

$$T_{\rm NR}=\left(\frac{2\pi}{\mu}\right)\frac{1}{\gamma+i\sqrt{2\mu E+i\epsilon}}\ ,$$ (12)

which depends, besides on the energy, only on the binding energy and the masses of the external particles. In particular one gets for the residue at the pole

$$g^{2}_{\rm NR}=\lim_{E\to-E_{B}}(E+E_{B})T_{\rm NR}=\frac{1}{d\Pi(E)/dE|_{E=-E_{B}}}=\frac{2\pi\gamma}{\mu^{2}}\ ,$$ (13)

also independent of the detailed dynamics that lead to the appearance of the pole. In fact, shallow bound states develop some universal properties—for a review see Ref. [57].

The discussion above applies to purely dynamically generated states. The equations were generalised by Weinberg [58] (inelastic channels were later included in Ref. [59]). The study starts from a two component wave function of a bound state interacting via some hamiltonian $\mathcal{H}$:

$$|\Psi\rangle=\left(\lambda|\psi_{0}\rangle\atop\chi(\mathbf{p})|h_{1}h_{2}\rangle\right)\qquad\mbox{solving}\qquad\hat{\mathcal{H}}|\Psi\rangle=E|\Psi\rangle,~{}~{}\hat{\mathcal{H}}=\left(\begin{array}[]{cc}\hat{H}_{c}&\hat{V}\\ \hat{V}&\hat{H}_{hh}^{0}\end{array}\right)\ .$$ (14)

Here $\psi_{0}$ ($h_{1}h_{2}$) is the compact (two hadron) component of the wave function. The terms in the hamiltonian include originally the interquark potential which also contains the mechanism of confinement, $\hat{H}_{c}$, the transition potential between the compact component and the continuum, $\hat{V}$, and the hamiltonian for the two-hadron system. The quantity of interest here is

$$\lambda^{2}=\langle\psi_{0}|\Psi\rangle^{2}\ ,$$ (15)

which is the probability to find the compact component of the wave function in the full wave function and is thus a direct measure of the composition of the wave function. In earlier works Weinberg demonstrated that the non-perturbative parts of the hadron-hadron interaction can be absorbed into the effective parameters of the formalism by a proper field redefinition  such that to leading order in a momentum expansion $\hat{H}_{hh}^{0}=p^{2}/(2\mu)$ [63, 64]. With this one finds for the two-hadron wave function

$$\chi(p)=\lambda\,\frac{f(p^{2})}{E-p^{2}/(2\mu)}\qquad\mbox{with}\qquad f(p^{2})=\langle h_{1}h_{2}|\hat{V}|\psi\rangle\ .$$ (16)

The normalisation condition for physical bound states then can be expressed as

$$1=\langle\Psi|\Psi\rangle=\lambda^{2}\left(1+\int\frac{d^{3}p}{(2\pi)^{3}}\frac{f^{2}(p^{2})}{(E_{B}+p^{2}/(2\mu))^{2}}\right)=\lambda^{2}\left(1+g_{0}^{2}\int\frac{d^{3}p}{(2\pi)^{3}}\frac{1}{(E_{B}+p^{2}/(2\mu))^{2}}+\mathcal{O}\left(\frac{\gamma}{\beta}\right)\right)=\lambda^{2}\left(1+\frac{g_{0}^{2}\mu^{2}}{2\pi\gamma}+\mathcal{O}\left(\frac{\gamma}{\beta}\right)\right)\ ,$$ (17)

where $g_{0}=f(0)$ and $\beta$ denotes the intrinsic momentum scale in the transition form factor $f(p^{2})$. This allows one to relate the effective coupling $g_{0}^{2}$ to the quantity of interest, $\lambda^{2}$,

$$g_{0}^{2}=\frac{2\pi\gamma}{\mu^{2}}\left(\frac{1}{\lambda^{2}}-1+\mathcal{O}\left(\frac{\gamma}{\beta}\right)\right)\ .$$ (18)

It is straightforward to calculate the scattering amplitude within the same formalism under the assumption that the scattering is dominated by a single pole. All it takes is to replace in Eq. (10) $C$ by $g_{0}^{2}/(E-M_{0})$. Now the scale dependence is to be absorbed into the bare mass $M_{0}$ via that renormalisation condition $E_{B}=-M_{0}+g_{0}^{2}\mu/(2\pi)(\Lambda-\gamma)$, where we used that the analytic continuation of the square root of the energy on the first sheet (here the focus is on bound states) gives the positive imaginary momentum. Then the scattering matrix reads [8]

$$T_{\rm NR}=\frac{g_{0}^{2}}{E+E_{B}+g_{0}^{2}\mu/(2\pi)(ip+\gamma)}\ .$$ (19)

For a pure molecule we have $\lambda\to 0$ and accordingly $g_{0}^{2}\to\infty$. In that limit Eq. (19) agrees with Eq. (12). Moreover, to get from the bare coupling provided in Eq. (18) to the residue at the pole, it needs to be multiplied with the $Z$-factor that happens to agree with $\lambda^{2}$. This gives

$$g_{\rm NR}(\lambda^{2})^{2}=\lambda^{2}g_{0}^{2}=\frac{2\pi\gamma}{\mu^{2}}\left(1-{\lambda^{2}}+\mathcal{O}\left(\frac{\gamma}{\beta}\right)\right)\ ,$$ (20)

which agrees to Eq. (13) for $\lambda^{2}=0$ when the range corrections are omitted. Eqs. (18) and (19) nicely illustrate what the Weinberg criterion really does: It measures the importance of the term non-analytic in the energy, $p=\sqrt{2\mu E}$, which appears since the two-hadron intermediate state can go on shell and is thus specific for the molecular component, relative to the terms analytic in energy that also appear for compact structures.

The effective range expansion (ERE) reads

$$T_{\rm NR}(E)=-\frac{2\pi}{\mu}\frac{1}{1/a+(r/2)p^{2}-ip}\ ,$$ (21)

with $a$ and $r$ for the scattering length and the effective range, respectively, and we used the particle physics sign convention for the scattering length. Matching Eq. (21) and Eq. (19) gives

$$\frac{1}{a}=-\frac{2\pi E_{B}}{\mu g_{0}^{2}}-\gamma\ ,\quad r=-\frac{2\pi}{g_{0}^{2}\mu^{2}}\qquad\Longrightarrow\qquad{a=-2\frac{1-\lambda^{2}}{2-\lambda^{2}}\left(\frac{1}{\gamma}\right)+{\cal O}\left(\frac{1}{\beta}\right)}\ ,\quad{r=-\frac{\lambda^{2}}{1-\lambda^{2}}\left(\frac{1}{\gamma}\right)+{\cal O}\left(\frac{1}{\beta}\right)}\ ,$$ (22)

where Eq. (18) was used to provide relations in terms of the probability $\lambda^{2}$. One thus finds that for a pure molecule, $\lambda^{2}=0$, $a=-1/\gamma+{\cal O}\left({1}/{\beta}\right)$ and $r={\cal O}\left({1}/{\beta}\right)$ and for a purely compact structure, $\lambda^{2}=1$, $a={\cal O}\left({1}/{\beta}\right)$ and $r\to-\infty$. While the scattering length is very sensitive to the actual binding energy, it is the effective range that is most sensitive to the binding mechanism and thus to the structure of the state as will be discussed below. Moreover, because of the range corrections indicated in Eqs. (22) one should not use $a$ and $r$ as input to extract $\lambda^{2}$ [65], but more use the relations to check if, within errors, the properties of a given system are consistent with e.g. a pure molecule ($\lambda^{2}=0$).

Figure 5: The scatteringlength-effective range plane: Shown is for which pairs of values what kind of pole emerges. The area left white and the hatched one are forbidden from causality—the former because the related poles are located in the complex plane of the first sheet, the latter because of the Wigner bound. The dotted line ($r=-a$) refers to a pole with a real part exactly at the threshold. The red star (blue square) shows a typical location of the pole of a bound state with a molecular structure (compact structure). The respective arrows show the characteristic trajectory of the poles as the interaction gets changed slightly. The figure is adapted from Ref. [66].

In Fig. 5 the red star and the blue square show the pole locations for bound states with a molecular and compact structure in a plane defined by scattering length and effective range, respectively. The figure also assigns regions in the $(r-a)$ plane to the types of poles, namely bound states, real valued poles on the first sheet of the complex energy plane, virtual states, real values poles on the second sheet, and resonances, complex valued poles in the complex plane of the second sheet. The probabilistic interpretation of $\lambda^{2}$ was derived from the normalisation condition of the bound state. Clearly it cannot be copied straightforwardly to virtual states and resonances, since their wave functions cannot be normalised. However, it is instructive to ask what happens, if the interaction strength that lead to the pertinent pole gets slightly weakened (e.g. by changing the quark mass). Since the effective range is controlled by the type of the binding interaction it will change little, however, the scattering length will change its sign (see, e.g., Refs. [67, 68] and [69] for detailed studies of light and singly heavy systems; the situation is potentially more complicated for doubly heavy systems as described below): For molecular structures it is the inverse scattering length that changes smoothly (a molecular bound state exactly at threshold is characterised by an infinite scattering length), while for compact structures it is the scattering length itself. As indicated in Fig. 5, the weakening of the interaction thus transforms a molecular bound state into a virtual state, but a compact state into a resonance. One thus needs to conclude that a virtual pole necessarily is generated from non-perturbative two-hadron interactions and cannot be generated from a compact state. There are a large number of works available in the literature that discuss generalizations of Weinberg’s criterion also to resonances [59, 66, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79], which, however, we cannot discuss in detail here.

The derivation provided so far was based on single channel scattering with zero range interactions only, however, crucial information is encoded in the effective range. In fact, it appears that a positive effective range is an unambiguous signature of a purely molecular state [58, 80, 81, 82, 83], fully in line with the general theorem that potential scattering in a single channel with purely attractive interactions necessarily has a positive effective range [84]. Unfortunately reversing this statement does not work: In general not even a sizeable negative effective range provides a unique signature of a compact structure, since coupled channel effects also induce a negative contribution to the effective range. For illustration we may start from generalising Eq. (19) to two channels to find [81]

$$T_{\rm NR}^{(ij)}=\frac{g_{0}^{(i)}g_{0}^{(j)}}{E+E_{B}+g_{0}^{(1)\,2}\mu_{1}/(2\pi)(ip_{1}+\gamma_{1})+g_{0}^{(2)\,2}\mu_{2}/(2\pi)(ip_{2}+\gamma_{2})}\ ,$$ (23)

where the on-shell momenta of the particles in channel $i$ are

$$p_{i}(E)=\sqrt{2\mu_{i}(E-\delta_{i})}\Theta(E-\delta_{i})+i\sqrt{2\mu_{i}(\delta_{i}-E)}\Theta(\delta_{i}-E)$$

using the channel specific reduced masses $\mu_{i}$ and as before $\gamma_{i}=p_{i}(-E_{B})$. If the total energy is measured relative to the lowest threshold, which we assign to channel 1, one gets $\delta_{1}=0$ and $\delta_{2}=m_{1}^{(2)}+m_{2}^{(2)}-m_{1}^{(1)}-m_{2}^{(1)}$. In principle the denominator of Eq. (23) could also contain additional inelasticities that we neglect here, for simplicity. Clearly, the higher channel provides a contribution to the effective range, which now reads [66, 80, 81]

$$r=-\frac{2\pi}{g_{0}^{(1)\,2}\mu_{1}^{2}}-\frac{g_{0}^{(2)\,2}\mu_{2}}{g_{0}^{(1)\,2}\mu_{1}^{2}}\sqrt{\frac{\mu_{2}}{2\delta_{2}}}+{\cal O}\left(\frac{1}{\beta}\right)\ .$$ (24)

The second term on the right hand side follows straightforwardly from expanding $p_{2}$ around $E=0$. It should be stressed that the sign of this term is fixed from unitarity. In other words, the effect of heavier channels on the effective range is necessarily negative and can thus mimic the presence of a compact component of a studied shallow resonance. In case of the $X(3872)$ or the $T_{cc}(3875)$ the distances between the two pertinent thresholds are with $\delta_{2}{=}m_{D^{+}}{+}m_{\bar{D}^{*\,-}}{-}m_{D^{0}}{-}m_{\bar{D}^{*\,0}}{=}8$ MeV and $\delta_{2}{=}m_{D^{+}}{+}m_{D^{*\,0}}{-}m_{D^{0}}{-}m_{D^{*\,+}}{=}1.4$ MeV, respectively, very small, since both emerge from the isospin violating mass differences between the charged and neutral $D^{(*)}$-mesons. Accordingly, the second term above is e.g. with $-1.4$ fm for the $X(3872)$ quite sizeable and should be subtracted before the Weinberg criterion is applied. In Ref. [81] it is argued that the recent analysis of the line shape of $X(3872)$ by LHCb [86] only allows for an extraction of a lower bound for $g_{0}^{(1)\,2}$ and accordingly one is to conclude from current line shape studies that the parameters of the $X(3872)$ are fully consistent with a pure molecule.

The underlying physics of the above mentioned range corrections is the exchange of mesons of a finite mass. Mathematically spoken these exchanges introduce a left-hand (energies smaller than 0) branch-point and with it a left-hand cut (lhc) into the amplitude, located at [87]

$$E_{\rm lhc}^{x}=\frac{1}{8\mu}\left[(\Delta M)^{2}-m_{x}^{2}\right]$$ (25)

for the exchange of a particle of mass $m_{x}$ and a mass difference of the external particles in the process of the emission of $\Delta M$. This branch-point introduces a non-analyticity into the amplitude that invalidates the ERE in the simple form provided in Eq. (21). The left-hand branch point is closest to the physical axis for light exchange particles. Accordingly, the leading lhc in case of nucleon-nucleon scattering (as long as we neglect the exchange of photons), where $\Delta M=0$, comes from the one-pion exchange and is located at $E_{\rm lhc}^{\pi}[NN]=-5$ MeV—in this case the ERE can be used to extract the pole of the deuteron located at $-2.2$ MeV. In case of $BB^{*}$ scattering, where $\Delta M{=}M_{B}^{*}{-}M_{B}{=}45$ MeV (in the process of a pion emission a $B$ gets converted into a $B^{*}$), we get $E_{\rm lhc}^{\pi}[BB^{*}]=-2$ MeV. For $DD^{*}$ scattering there is no lhc, since here $\Delta M{=}M_{D}^{*}{-}M_{D}{=}140.6$ MeV${>m_{\pi}}$ and a positive value for $E_{\rm lhc}^{x}$ indicates the presence of a three-body cut instead of a lhc. However, for slightly larger than physical pion masses, as is studied in lattice QCD, the left-hand cut is located very close to the physical axis. For examply, for $m_{\pi}=280$ MeV and 1927 and 2049 MeV for the masses of the $D$ and the $D^{*}$ meson as used in Ref. [88], the left-hand branch point is located at $-8$ MeV while the pole extracted from the lattice data using the ERE is located at $-10$ MeV. Thus, in this case the ERE in its original formulation should not be employed—see Ref. [89]. Moreover, to extract phase shifts from the lattice energy levels the Lüscher method was employed that also calls for a modification in the presence of left-hand cuts [90, 91, 92]. Moreover, the pole trajectories in this case look a lot more complicated than what is discussed in Refs. [67, 68, 69] as detailed in Refs. [93, 94].

Figure 6: The relation between $Y(4230)\to\pi Z_{c}(3900)$ and $Y(4230)\to\gamma X(3872)$ in leading order EFT und der assumption that all external charmonium like states are hadronic molecules. Fig. from Ref. [11].

As demonstrated above, important information on the nature of the state is encoded in the effective coupling $g_{0}^{2}$ or, more concretely, the related residue, $g_{\rm NR}(\lambda^{2})^{2}$, which gets maximal for $\lambda^{2}=0$. Having that said it becomes clear that only those observables where one is sensitive to the mentioned residue can be sensitive to the molecular component. However, this is not always given. For example, the decays $X(3872)\to\gamma\psi$, where $\psi$ is either $\psi(2S)$ or $J/\psi$, can either go via a $D^{*}\bar{D}\to\gamma D^{(*)}\bar{D}^{(*)}\to\gamma J/\psi$ triangle diagram, which scales with $g_{\rm NR}(\lambda^{2})$, or a contact transition, which does not. The latter type of diagram is necessary to absorb the divergence of the former. Thus, the total rate of this transition cannot be sensitive to the molecular component of the $X(3872)$ [95] in a model independent way, contrary to what is claimed, e.g., in Refs. [96, 97]. However, clearly these radiative decays are excellent observables to test model predictions [98, 96, 99, 100, 101, 102] (see also the compilation provided in Ref. [103]). On the other hand for some other decays non-trivial predictions are possible. For example, if $Y(4230)$ and $Z_{c}(3900)$ are both molecular made of $D_{1}\bar{D}$ and $D^{*}\bar{D}$, respectively, there is no leading order counter term for the transition $Y(4230)\to\pi Z_{c}(3900)$, which starts at one loop level [104]—see left diagram in Fig. 6. Moreover, this transition is enhanced by a nearby triangle singularity [105]. If, in addition, also the $X(3872)$ is a $D^{*}\bar{D}$ molecular state, then the same topology contributes to $Y(4230)\to\gamma X(3872)$ (see right diagram of Fig. 6) such that its rate could be predicted [106]. The prediction was confirmed shortly after by BESIII [107].

Figure 7: Current spectrum of charmonium and charmonium like states. Solid lines show states well established while dashed lines those where confirmation is still to happen. Blue lines relate to states showing properties in line with the quark model (using the results of Ref. [108] as a guidance), red lines those with unconventional properties. To distinguish isovector states a $+$ is attached to the name. Thin dashed lines indicate the thresholds for the particle pairs put on the left (charge conjugation is understood to be included). The thick, solid lines show the lowest thresholds, where two narrow open charm states can combine to the given quantum numbers.

After the quite general discussion about hadronic molecules we now come to the direct implications of a molecular structure for doubly heavy systems. For the rest of this chapter we focus on the hypothesis that the states studied are in fact purely molecular and ask what imprint that has on observables. We start with the spectrum. First of all it is important to note that only narrow hadrons can form hadronic molecules, since typically the widths of the constituents set a lower limit to the width of the molecule [109]—the width can get smaller than that of the constituents only, if the reduction in phase space provided by the bound state mass being smaller that the nominal threshold of the constituents is relevant as is the case for the $T_{cc}$. Another view on this situation is to acknowledge that the time scale of the formation of the molecule needs to be much shorter than the life-time of the constituents [110]. Thus we have at our disposal as building blocks for hadronic molecules the spin doublets $\{D,D^{*}\}$, with negative parity and negligible widths, and $\{D_{1}(2420),D_{2}(2460)\}$ with positive parity and widths below 50 MeV. All 4 states fill a flavor $[\bar{3}]$ multiplet. In contrast to the listed narrow states, the doublet $\{D_{0}(2300),D_{1}(2430)\}$ having widths of the order of 300 MeV, is not expected to form hadronic molecules.

Since the centrifugal barrier acts like a repulsive force, in general hadronic molecules should typically appear in $S$-waves. Accordingly, the quantum numbers of the constituents dictate those of the composite state. In Fig. 7 both is shown the thresholds for the various hidden charm channels (as the thin dashed lines) as well as the lowest threshold where the given quantum numbers can be reached in an $S$-wave by combining the quantum numbers of the constituents [15]. It is interesting to observe that with one exception—the $T_{c\bar{c}0}(4240)^{+}$, that still awaits confirmation—all states with unconventional properties observed so far appear either just below or above the thick black line, as expected for hadronic molecules. Moreover, from these considerations it follows that, if the $Y(4230)$ is a $D_{1}\bar{D}$ molecular state [111, 104, 112], its lightest spin partner state with $J^{P}=0^{-}$ must be located near the $D_{1}\bar{D}^{*}$ threshold and thus be 140 MeV heavier than the $Y(4230)$. This prediction was confirmed in a microscopic calculation [113] that even puts a state with the exotic quantum numbers $J^{PC}=0^{--}$ in this mass range. The very same calculation also identifies $\psi(4230)$, $\psi(4360)$, and $\psi(4415)$ as hadronic molecules with $J^{PC}=1^{--}$ and binding energies between about 50 and 70 MeV in the channels $D_{1}\bar{D}$, $D_{1}\bar{D}^{*}$ and $D_{2}\bar{D}^{*}$, respectively.

A state that does not fit into the classification as $S$-wave hadronic molecules is the $Z_{c}(4430)$, since it sits in the mass range of the $D^{*}\bar{D}_{2}$ threshold, but has positive parity. A possible explanation for a two-hadron structure of it could be a $P$-wave $D^{*}\bar{D}_{1}$ state [114]. This explanation calls for assigning the $Y(4390)$ as its $S$-wave companion. If this explanation were correct, there should be a signature of the $Z_{c}(4430)$ in the $D^{*}D^{*}\pi$ subsystem of e.g. $B\to KD^{*}D^{*}\pi$ (so far the $Z_{c}(4430)$ showed up only in $B\to KZ_{c}(4430)$). While the model study of Ref. [115] does not confirm the mentioned assignment, it finds a series of isoscalar $P$-wave states. In Ref. [116] another meson exchange model is employed to connect a predicted $P$-wave $D\bar{D}^{*}$ resonance with the well established $X(3872)$ and $Z_{c}(3900)$ as well as the $T_{cc}(3875)$ all having a mass very close to the $D\bar{D}^{*}$ threshold.

Figure 8: The reactions $e^{+}e^{-}\to J/\psi\pi\pi$ (left panel) and $e^{+}e^{-}\to D^{*}\bar{D}\pi$ (right panel) calculated in the molecular model. The red solid line shows the results of the full model of Ref. [117], while for the blue dashed line the effect of the $\psi(4160)$ was switched of. The data are from Refs. [18] and [19] for the left and right panel, respectively.

The key characteristic of a hadronic molecule is that it has a strong coupling to its constituents. Therefore, the nearby pole must lead to some imprint of the pertinent continuum threshold in observables. This is illustrated in Fig. 8 on the example of two decay channels of $Y(4230)$. The lines are from a model study that includes the $D_{1}\bar{D}$ molecular dynamics—in particular the effect of the $D_{1}\bar{D}$ channel is included fully dynamically [117]. Moreover, the data (also in other channels) called for the inclusion of not only the $Y(4230)$ but also the $\psi(4160)$. With these ingredients it was possible to describe almost all available data for the $Y(4230)$ with a consistent model. Especially, no extra exotic state near 4320 MeV is necessary to describe the data, contrary to the analyses of e.g. Refs. [18, 118], since the highly asymmetric line shape of the $J/\psi\pi\pi$ channel is generated from the $D_{1}\bar{D}$ cut. The line-shape that emerged for the $D^{*}\bar{D}\pi$ channel is characteristic for the decay for a hadronic molecule with an unstable constituent having a width of the same order of magnitude as the binding energy [119, 120].

Typically the interaction of two heavy mesons is in principle modelled analogously to the nucleon-nucleon interaction, reviewed in Ref. [56]. However, since the data situation is a lot worse than for the two-nucleon system, there is no consensus yet on the most relevant contribution to the binding potential. While many follow Ref. [4] which suggest that vector meson exchanges should provide the bulk of the binding [121, 122], others follow Ref. [5] assuming the one-pion exchange as the prime contribution to the binding [123]. Other works include both mechanism together with others [124, 125, 126]—clearly the list of references is not exhaustive.

In addition to model calculations there are effective field theory (EFT) studies for doubly heavy molecular states. It should be clear that they suffer from less predictive power than model predictions. For example

• meson-meson and meson-antimeson scattering are not related in an non-relativistic EFT,

• the short ranged operators for different total isospin channels are not related and thus the existence of e.g. an isovector partner of a molecular $X(3872)$ can be deduced only from an analysis of high quality data [127] and

• it is not even possible to exploit heavy quark flavor symmetry for doubly heavy systems [128].

However, EFTs allow one to fully and systematically exploiting the implications of the symmetries of the underlying theory (here QCD) on the hadronic observables and thus allow for model independent insights. There are in general three different classes of approaches available in the literature, namely pionless EFT [129, 130], EFT with perturbative pions [131, 132, 133, 134, 135, 136], and EFT with non-perturbative pions [137, 138, 139, 140, 141, 142, 143, 144]. Again, also those find their analogs in studies of the nucleon-nucleon interaction [56].

Figure 9: Effect of spin symmetry violation on the predicted spin multiplet structure for the spin partner states of $Z_{b}(10650)$ and $Z_{b}(10610)$. Fig. from Ref. [11].

One prediction of those effective field theories is that, if the bottomonium like exotic states $Z_{b}(10610)$ and $Z_{b}(10650)$ with $I(J^{P})=1(1^{+})$ are both hadronic molecules with a decomposition $B\bar{B}^{*}$ and $B^{*}\bar{B}^{*}$, respectively, then there are necessarily two multiplets of spin partner states at least in the heavy mass limit, one containing the $Z_{b}(10650)$ and a scalar state, which disappears as soon as finite mass corrections are included for the $B$ mesons, and one containing the $Z_{b}(10650)$ together with 3 $W_{bJ}$ states, $J=0,1,2$ [133, 145, 142]. As spin symmetry violation is introduced into the system, the degeneracy within the multiplets is lifted and the different states get attached to certain thresholds—see Fig. 9 (while one of the scalar states gets unbound). What is also clear from the figure is that for the spectrum the pionless and the pionfull theory give very similar results, however, the presence of pions gives the $W_{b2}$ a significant width. It should be stressed that the tight connection between the spin symmetry violation of the hadronic molecules and that of the constituents is a unique prediction of molecular models.

4 Hidden-charm pentaquarks

Multiquarks are expected not only in the meson sector but also in the baryon sector, where the simplest structures are pentaquarks. The observation of a broad $P_{c\bar{c}}(4380)$ and a narrow $P_{c\bar{c}}(4450)$ [146] in 2015 at LHCb was the first undebated observation of pentaquarks. They decay strongly into $J/\psi p$. Since the $J/\psi$ cannot be produced in the process of the decay, the states contain at least five quarks, $c\bar{c}uud$. The data were analyses further in Refs. [147, 148] which supported the existence of two states. A later analysis based on an order of magnitude larger data sample, shows that the $P_{c\bar{c}}(4450)$ structure consists of two narrow overlapping peaks, $P_{c\bar{c}}(4440)$ and $P_{c\bar{c}}(4457)$, and a third narrow peak $P_{c\bar{c}}(4312)$ appears [149]. However, there is no longer clear evidence for the broad $P_{c\bar{c}}(4380)$. As in the meson sector, these discoveries were followed by numerous theoretical interpretations of the nature of the pentaquarks, including hadronic molecules [150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 121, 179], compact pentaquark states [180, 181, 182, 183, 184, 185, 175], hadro-charmonia [186, 187, 188], and cusp effects [175].

The proximity of the $\Sigma_{c}\bar{D}^{(*)}$ thresholds to these narrow pentaquark structures suggests that the corresponding two-hadron states play an important role in the dynamics of the pentaquarks, suggesting an interpretation of their structure as hadronic molecules. In the most common hadronic molecular picture, the $P_{c\bar{c}}(4312)$ is an $S$-wave $\Sigma_{c}\bar{D}$ bound state, while the $P_{c\bar{c}}(4440)$ and $P_{c\bar{c}}(4457)$ are bound states of $\Sigma_{c}\bar{D}^{*}$ with different spin structures [153, 158, 164, 172, 178]. The origin of the peak from the $P_{c\bar{c}}(4312)$ is attributed to a virtual state of $\Sigma_{c}\bar{D}$ in Ref. [189] based on amplitude analysis, which only fits data around the $\Sigma_{c}\bar{D}$ threshold. In Ref. [175], the final state interactions are constructed based on a $K$ matrix containing the $J/\psi p$-$\Sigma_{c}\bar{D}$-$\Sigma_{c}\bar{D}^{*}$ channels. The analysis suggests that $P_{c\bar{c}}(4312)$ is a $\Sigma_{c}\bar{D}$ molecule, while the $P_{c\bar{c}}(4440)$ could be a compact pentaquark state, and the $P_{c\bar{c}}(4457)$ could be caused by the cusp effect.

While the proximity of the narrow $P_{c\bar{c}}$ peaks to the $\Sigma_{c}\bar{D}^{(*)}$ thresholds makes the molecular interpretation for the states natural, at least some peaks in the $J/\psi p$ mass distributions can also be generated by triangle singularities [193, 194, 195, 196, 149, 191]. A triangle singularity arises when all intermediate particles in a triangle loop are (almost) on the mass shell and the particles move collinearly. Therefore, its location is quite sensitive to the masses and widths of the particles involved [191]. The potential triangle singularities have been discussed in Ref. [149] for the three $P_{c\bar{c}}$ structures. Considering the realistic widths of the exchanged resonances, the $P_{c\bar{c}}(4312)$ and $P_{c\bar{c}}(4440)$ structures are unlikely to be caused by triangle singularities. However, the $P_{c\bar{c}}(4457)$ structure could in principle be generated by a triangle diagram with $D_{s1}^{*}(2860),\Lambda_{c}(2595)$ and $\bar{D}^{*0}$ in the intermediate state.

Under the assumption that the observed pentaquarks are molecular in nature, heavy quark spin symmetry allows one to predict spin partner states. Combining the light quark spins of the $\Sigma_{c}$, $j_{\rm light}^{P}=1^{+}$, and the ground state $D$ mesons, $j_{\rm light}^{P}=(1/2)^{+}$, allows for two distinct total angular momenta of the light quarks, namely $1/2$ and $3/2$. Accordingly the dynamics in the $\Sigma_{c}^{(*)}\bar{D}^{(*)}$ channels is controlled by two contact terms [153]. It turns out that heavy-quark spin symmetry (HQSS) predicts seven $P_{c\bar{c}}$ states, divided into two heavy quark spin multiplets. Three of these seven correspond to those reported by LHCb [197, 198, 153, 172]: While the $P_{c\bar{c}}(4312)$ is unambiguously assigned to the $J^{P}=\frac{1}{2}^{-}$ $\Sigma_{c}\bar{D}$ bound state, there are two possible spin structures for the $P_{c\bar{c}}(4440)$ and $P_{c\bar{c}}(4457)$ identified as the $\Sigma_{c}\bar{D}^{*}$ bound states, namely $J^{P}=\frac{1}{2}^{-}$ and $J^{P}=\frac{3}{2}^{-}$ [155, 153, 172], and their spin assignment is not uniquely determined by HQSS alone [151, 199, 154, 167, 172]. However, once the one-pion exchange (OPE) potentials were included, only one solution could be found, suggesting that $P_{c\bar{c}}(4440)$ and $P_{c\bar{c}}(4457)$ couple dominantly to the $\Sigma_{c}\bar{D}^{*}$ with quantum numbers $J^{P}=\frac{3}{2}^{-}$ and $\frac{1}{2}^{-}$, respectively [172]. In addition, there is evidence for an additional narrow state, also required by HQSS, around $4.38$ GeV was found in the data with $J^{P}=\frac{3}{2}^{-}$, which couples dominantly to the $\Sigma_{c}^{*}\bar{D}$ (see also Ref. [157, 178]).

While there is no evidence in the data yet for the three highlying states, they have to exist, if the pentaquarks observed are indeed of molecular nature. In addition, all the different states need to prominently decay into the channels that form the molecular states (that is $\Sigma_{c}\bar{D}$ for the $P_{c\bar{c}}(4312)$). The pertinent predictions are provided in Ref. [200]. The relevant data are expected to be published in the coming years.

5 All heavy tetraquarks

While most of the experimental information is available for hadrons containing two heavy constituents, the LHCb measurements of di-$J/\psi$ production in proton-proton collisions at center-of-mass (c.m.) energies 7, 8 and 13 TeV revealed a new, potentially rich class of exotic states with four charmed (anti-)quarks [201]. In fact, the measured line shape has a non-trivial shape that deviates significantly from the phase space distribution as well as from the exponential behaviour predicted from perturbative QCD for single- and double-parton scattering. The attention of both the experimental and theoretical communities was mainly drawn to the statistically significant peak structures observed in the energy range from 6.5 GeV to about 7.2 GeV. In particular, the most prominent structure was named $T_{cc\bar{c}\bar{c}}(6900)$ (also known as $X(6900)$) [202]. A fully-charmed compact tetraquark resonance is the most natural candidate to explain the structure. However, most of the theoretical studies give $cc\bar{c}\bar{c}$ ground states at a mass significantly lower than 6.9 GeV [203, 204, 205, 206, 207, 208, 209, 210, 211, 212]. Thus, one expects lower states, if there is a $cc\bar{c}\bar{c}$ resonance with a mass around 6.9 GeV. Due to a smaller phase space, such lighter states are expected to have smaller widths. However, there are no obvious narrower peaks in the reported double-$J/\psi p$ spectrum.

When the data are analysed using a coupled-channel scattering formalism with vector charmonium pairs [213], the number and locations of the poles in the 6900 MeV mass region appear to be rather poorly determined and strongly dependent on the model ingredients. Nevertheless, if this approach does indeed capture the relevant dynamics, the data suggest the existence of a pole near the di-$J/\psi$ threshold in the double-$J/\psi$ production amplitude. This state was named $X(6200)$ (or $T_{cc\bar{c}\bar{c}}(6200)$ according to the new naming scheme for exotic particles promoted by the Particle Data Group [202]). This finding was confirmed by the calculation in Ref. [214] and more recently in Ref. [215]. Due to the suppression of the signal near the threshold caused by the phase space factor, this pole cannot show up as a pronounced peak structure above the threshold in the double $J/\psi$ line shape. Instead, it can only be unambiguously identified by a comprehensive pole search in the coupled-channel scattering amplitudes. Meanwhile, a steep rise of the line shape just above the threshold provides evidence for the existence of such a near-threshold pole [190]. As can be seen from the analysis in Ref. [213], the behaviour of the signal just above the di-$J/\psi$ threshold does indeed call for the existence of the $X(6200)$ pole.

Recently, data on the double $J/\psi$ production in $pp$ collisions have also arrived from two other LHC collaborations, namely CMS [216] and ATLAS [217]. Remarkably, in both cases the data description improved after the inclusion of an auxiliary Breit-Wigner resonance centred just above the production threshold. The need for its inclusion in the fitting function supports the existence of the $X(6200)$ pole in the amplitude [218].

6 Single heavy tetraquarks

Since their discovery in 2003 the lightest open charm positive parity states containing strangeness remained largely a mystery—especially in light of the seemingly expected properties of their non-strange partner states. In the past, attempts were made to explain the low lying $D_{s}$ states like $c\bar{s}$ mesons [219, 220, 221, 222, 223], chiral partners of the ground state $D_{s}$ and $D_{s}^{*}$ mesons [224, 225], compact $[cq][\bar{s}\bar{q}]$ tetraquark states [26], mixing of the $c\bar{s}$ and tetraquarks [226], a $D\pi$ atom for the $D_{s0}^{*}(2317)$ [227], and $D^{(*)}K$ hadronic molecules [228, 229, 230, 231, 232, 233]. The experimental data show three features that need to be understood:

1. The $D_{s}$ states are too light: Both, $D_{s0}^{*}(2317)$ and $D_{s1}(2460)$ are much lighter than the quark model expectations for the lowest scalar and axial-vector $c\bar{s}$.

2. Fine-tuning: One has $M_{D_{s1}(2460)}-M_{D_{s0}^{*}(2317)}\simeq M_{D^{*\pm}}-M_{D^{\pm}}$ within 2 MeV.

3. Mass hierarchy: One finds $M_{D_{0}^{*}(2300)}\sim M_{D_{s0}^{*}(2317)}$ and $M_{D_{1}(2430)}\sim M_{D_{s1}(2460)}$, although usually adding a strange quark leads to an increase in mass of about 150-200 MeV.

Items (1) and (2) could be understood in a compact tetraquark picture [234], however, leaving the third one unexplained. On the other hand, all these find a simultaneous natural explanation, if the lowest positive-parity charmed mesons are interpreted as hadronic molecules. In this case the flavor structure of this family of states is governed by the one that emerges from the flavor decomposition representing the scattering of Goldstone bosons off $D$ mesons, which may be expressed as [235]

$$[\mathbf{\bar{3}}]\otimes[\mathbf{8}]=[\mathbf{\bar{3}}]\oplus[\mathbf{6}]\oplus[{\mathbf{\overline{15}}}]\ ,$$ (26)

where the multiplets on the left refer to the $D$ mesons and the light pseudoscalars, respectively. Note that we do not include scattering of the ninth pseudoscalar, the $\eta^{\prime}$, here, since due to the action of the $U(1)_{A}$ anomaly it cannot be regarded as a Goldstone boson. Non-strange isospin 1/2 multiplets appear in all three irreducible representations, however, chiral symmetry constraints dictate that only the $[\mathbf{\bar{3}}]$ and the $[\mathbf{6}]$ are attractive [235]. In particular, in this case the $D_{0}^{*}(2300)$ is interpreted as emerging from the interplay of two poles, one at 2105 MeV and one at 2451 MeV, with the lower one being a member of the same SU(3) multiplet as the $D_{s0}^{*}(2317)$, the $[\mathbf{\bar{3}}]$, where the attraction is the strongest. The state at 2451 MeV is a member of the $[\mathbf{6}]$ representation of SU(3), where the interaction is weaker than in the $[\mathbf{\bar{3}}]$, but still sufficiently strong to generate a pole at physical quark masses sufficiently close to the physical axis to show an impact on observables [231, 236, 237, 238]. It should be stressed that two-pole structures are occurring in various systems, since the compared to the naive quark model enlarged multiplet structure alluded to in Eq. (26) appears analogously in various systems. For a recent review see Ref. [239]. For the singly heavy, positive parity open flavor states chiral symmetry constraints impose that the interaction of the particles in the $[{\mathbf{\overline{15}}}]$, is repulsive and thus no state should be found in this channel.

Thus, a crucial test of this interpretation is connected to the existence of the $\mathbf{[6]}$ state with a pole located at 2451 MeV and the absence of a pole in the $[{\mathbf{\overline{15}}}]$. In contrast to this, quark model states with a quark composition $c\bar{u}$ or $c\bar{d}$ can appear only in the flavor $[\mathbf{\bar{3}}]$ representation.

One way to unravel the SU(3) structure underlying the spectrum of the lightest open charm scalar states is to unambiguously establish the existence of the state in the $\mathbf{[6]}$, by determining whether it appears as a near threshold pole in the case when the Goldstone Boson mass (i.e. the pion mass) is even larger and near the SU(3) symmetric point, as predicted by unitarized chiral perturbation theory [240]. This needs to be accompanied by the absence of a state in the $[{\mathbf{\overline{15}}}]$. First results indicated that, indeed, the $\mathbf{[\overline{15}]}$ state is repulsive and the $\mathbf{[6]}$ state is attractive, thus providing strong evidence for this state’s molecular nature [241]. This finding was confirmed recently [242] by a detailed Lüscher analysis imposing similar quark masses.

While these findings look like a strong support for a molecular structure of the mentioned states, it remains to be studied, what e.g. the compact tetraquark picture predicts for this system. In particular, in Ref. [243] it is claimed that only scalar diquarks should contribute to the formation of the mentioned states. Then it is indeed a consequence of the Pauli principle, that also compact tetraquarks only appear in the flavor $[\mathbf{\bar{3}}]$ and $[\mathbf{6}]$ representations and not in the $[{\mathbf{\overline{15}}}]$. Further experimental and theoretical studies are necessary to reveal the nature of the lightest positive parity open flavor states.

7 Candidates for light multiquarks

The emerging multiquark states in the light quark sector like the light scalar mesons $f_{0}(500)$, $f_{0}(980)$, $K_{0}^{*}(700)$ and $a_{0}(980)$ and the $\Lambda(1405)$ where already discussed in some detail in other sections of this text. We therefore do not repeat those discussions here but refer the readers to the corresponding sections in this encyclopaedia.

Figure 10: Illustration of the decomposition of tetraquark states into different components within the Dyson-Schwinger approach. The molecular component, the hadroquarkonium component and the compact tetraquark component are shown as diagram (a), (b) and (c), respectively. The figure is from Ref. [244].

(a)                             (b)                             (c)

8 Closing discussion on multiquark states

The sections above focused on the observable implications of various possible structures if they were realised in an isolated form. These are the scenarios that most of the literature is so far dealing with. However, before closing this chapter, recent approaches that allow one to study the interplay of different scenarios should be mentioned.

Phenomenologically the interplay of conventional mesons with multiquark states was studied in various works, however, the different works do not draw a clear picture. While e.g. Refs. [245, 246] call for a significant mixing of molecular and conventional components, Refs. [247, 248] claim, that in a limit of large coupling of conventional states to the continuum, the molecular structures that appear as collective phenomena of the whole tower of quark model states decouple from the conventional state.

One possible route with closer connection to QCD to access this field is to employ functional methods. Those were already introduced in other chapters of this encyclopedia, and applied to four-quark structures in Ref. [249, 244, 250]. The emerging ansatz for the four-body equations is shown in Fig. 10. In principle the method allows one to investigate the relative importance of the different components in the wave function of the exotic studied. The underlying interaction is already fixed from other studies. The technical problem here is that the equations can be solved with well established techniques in the space like regime, however, there are still some issues to overcome for a controlled continuation in the time-like regime, especially in the presence of various continuum thresholds.

Another very promising ansatz is the Born-Oppenheimer Effective Field Theory (BOEFT) [251, 252, 253, 254]. This effective field theory inherited from molecular physics uses from static quark-quark or quark-antiquark potentials as input for tailor made coupled-channel Schrödinger equations. A first exploratory study of the $X(3872)$ and the $T_{cc}(3875)$ can be found in Ref. [255]. This effective field theory holds the promise to combine hadronic and quark-gluon degrees of freedom in a field theoretically sound set up. It still needs to be see how the subtle analytic structure of e.g. the one pion exchange amplitude discussed in the previous section can be embedded into the formalism.

9 Conclusions

Given the fast developments both theoretically and experimentally, there is reason to believe that the nature of the recently discovered QCD exotics will be clarified within the next decade. Those insights will provide an additional important step to understand the inner workings of QCD in the non-perturbative regime.

[Acknowledgments]C.H. acknowledges the support from the CAS President’s International Fellowship Initiative (PIFI) (Grant No. 2025PD0087).