Is X(2370) a Glueball? What the Missing Decay Shows

A missing decay can test the X(2370) glueball hypothesis only when a competing description predicts that decay at a detectable rate. BESIII found no significant X(2370) signal in a selected K*(892)K̄ route and used the suppression in its flavour-singlet argument.

The issue is what follows from that limit. It supports a particular symmetry interpretation, but flavour-singlet behaviour does not uniquely identify a glueball. Production models, possible mixing and alternative descriptions must be compared with the same data.

The resonance is observed; its composition is the question

X(2370) has appeared in several BESIII final states. Its quantum numbers are J^PC = 0⁻⁺. The newer mass combination gives 2,359⁺¹³₋₁₄ MeV/c², while the total width is 170⁺⁴⁴₋₂₉ MeV. Our report on what BESIII measured explains the experimental sequence and the earlier 2024 fit.

A resonance fit establishes properties such as mass, width and quantum numbers within an analysis. It does not directly label the state’s quark and gluon content. Conventional mesons, hybrids and mixed configurations can share the same quantum numbers.

Radiative J/ψ decays provide a useful search environment because charm–anticharm annihilation produces a photon and a gluon-rich hadronic system. That environment favours the investigation of gluonic states, while also producing other hadrons. “Made in a gluon-rich process” is a reason to investigate the candidate, rather than a composition measurement.

What the K*(892) search actually selected

The July 2026 BESIII preprint used approximately ten billion J/ψ events. It examined J/ψ → γK_S⁰K_S⁰π⁰, then selected events with a K_S⁰π⁰ combination consistent with K*(892)⁰.

This searches for X(2370) → K*(892)⁰K̄⁰ plus the charge-conjugate process, followed by the neutral reconstructed final state. The inclusive sample contains an X(2370) structure; the K*-selected sample contains no significant X(2370) signal.

That is different from saying the resonance never decays. It appears through other routes into the same final-state particles. The search tests a particular intermediate decay path.

A null result also does not establish an exactly zero branching fraction. The collaboration fits the data, assesses uncertainties and reports an upper limit for the searched process.

Why 2.7 × 10⁻⁶ is a product branching-fraction limit

The bound applies to the probability of producing X(2370) in a J/ψ radiative decay multiplied by the probability of the searched reconstructed cascade:

B(J/ψ → γX) × B(X → K*⁰K̄⁰ + c.c. → K_S⁰K_S⁰π⁰) < 2.7 × 10⁻⁶.

Here, X denotes X(2370), K* denotes K*(892), and “c.c.” includes the charge-conjugate process. The limit is at 90% confidence level. It includes the specified reconstruction chain; it is not simply B(X → K*K̄).

The paper also bounds the fraction of the measured K_S⁰K_S⁰π⁰ decay proceeding through this route: R < 0.081 at 90% confidence level. The production factor cancels in that ratio, but the specified neutral final state remains.

A confidence limit is a result of a statistical procedure under its assumptions. It should not be translated into a 90% probability that the candidate is a glueball. Our guide to significance and the hypotheses a test addresses explains the broader distinction.

Why flavour-singlet behaviour does not identify the constituents

Flavour symmetry concerns how a state transforms under rotations among quark flavours such as up, down and strange. A flavour singlet is invariant under those transformations in the symmetry description.

Purely gluonic configurations carry no valence quark flavour. They are natural singlets. But a suitably combined quark–antiquark configuration, or a hybrid configuration, can also be a singlet. “Flavour singlet” and “contains no quarks” are not interchangeable statements.

BESIII invokes a selection argument for the 0⁻⁺ singlet’s K*K̄ decay. The observed suppression supports that line of reasoning. Its strength as a glueball test then depends on how other models treat the same channel.

Exact equal branching fractions into all quark flavours are not implied. Quark masses, phase space, symmetry breaking and mixing affect physical decay rates. The question is whether a specified model accounts for the pattern after those effects are included.

Seven inputs to the glueball interpretation

BESIII’s cumulative argument brings together inputs with different evidential roles. The ledger separates direct results from derived quantities and model comparisons.

X(2370): seven inputs, different kinds of evidence

A measurement, a limit and a composition inference do different jobs. No row supplies an automatic exclusion badge.

Mass

Measured2026 combination: 2,359⁺¹³₋₁₄ MeV/c². Compatibility with lattice masses does not identify constituents.

Quantum numbers

MeasuredJ^PC = 0⁻⁺. Appropriate for a pseudoscalar glueball; shared by other hadrons.

K*K̄ suppression

Upper limit → interpretationRadiative J/ψ → γX production × neutral K*K̄ cascade to K_S⁰K_S⁰π⁰: < 2.7 × 10⁻⁶, 90% CL. Supports the flavour-singlet argument.

Decay pattern

Channels → model comparisonSeveral three-meson modes are reported. Similarity to η_c is an interpretation of their pattern.

Partial widths

DerivedChannel widths require branching information. The combined 170 MeV total width is a different quantity.

γω and γφ modes

Suppression → interpretationBESIII cites these channels in its composition argument. Predictions depend on couplings and mixing.

Radiative production

Model comparisonPure-glue and mixed-state predictions differ. Extracting the production rate requires decay assumptions.

The cumulative glueball case must be tested against alternatives with the same inputs and uncertainties.

Why the production rate is a disputed test

Gui and colleagues’ quenched-lattice calculation predicted a pure pseudoscalar-glueball radiative-production branching fraction of (2.31 ± 0.90) × 10⁻⁴. Its assumptions define what that comparison tests.

In a peer-reviewed 2022 analysis, first posted in 2021, Sun and colleagues inferred a rate roughly an order of magnitude larger from the then-available decay spectra. They argued that this disfavoured the glueball interpretation. The extraction used a decay model; it was not a direct count of all X(2370) production.

BESIII’s 2026 discussion favours a gluonic interpretation and also challenges a decay-ratio consequence of that earlier model using newer channel information. The older discrepancy should therefore be described with its dataset and assumptions, rather than presented as a final contemporary exclusion.

Chen and colleagues’ 2026 preprint explores whether charmonium mixing could enhance the rate. A small admixture can have a large production effect in that picture. Compatibility under a mixing model still does not measure a unique glueball percentage.

The hybrid alternative remains a proposal to test

A hybrid meson combines a quark–antiquark component with an excited gluonic degree of freedom. Such states arise from the same low-energy QCD dynamics that also permit glueballs. Our confinement explainer gives the broader strong-interaction context.

Lu’s August 2026 preprint tests an exploratory flavour-singlet radial-hybrid model. It can suppress the K*K̄ route and accommodate aspects of the current data under its assumptions. Its comparison with a mixed-glueball description remains conditional on production inputs and uncertainty choices.

The study uses a surrogate decay model, rather than a complete first-principles prediction. Its compatibility is therefore not an observation of a hybrid, nor proof that every other model works equally well. It demonstrates a specific ambiguity worth investigating.

Conversely, the null decay result should not be turned into a blanket statement that every ordinary quark–antiquark configuration is excluded. Each proposed state needs a quantitative prediction and a comparison with the actual limit.

Which measurements could distinguish the descriptions?

Useful future tests should target predictions on which the models differ, rather than collect more examples of properties they share.

One route is a coherent amplitude analysis. A three-meson final state can be reached through several intermediate resonances. Their contributions interfere, so event totals alone do not determine the relative amplitudes and phases.

Lu’s study highlights channels such as f₀(980)η, a₀(980)π and K₀*(1430)K̄ as potentially informative under its forecast assumptions. Those projections are theoretical targets, rather than achieved experimental separations.

Production information, correlated decay measurements and lattice calculations including dynamical quarks would also improve the comparisons. Our discussion of simulating gauge-theory dynamics explains why computing strong-interaction behaviour is difficult, while distinguishing simplified models from full QCD.

A new result would become discriminating when competing descriptions make reliably different predictions and the analysis treats their experimental and theoretical uncertainties consistently.

How to state the glueball claim today

The strongest defensible distinction is between the observed resonance and the proposed dominant constituent. BESIII has assembled a cumulative case for a pseudoscalar-glueball contribution. The K*K̄ suppression adds a test of flavour behaviour, while production and decay models determine how uniquely the full pattern selects that interpretation.

The 2024 quantum-number paper and the 2026 combined-channel paper are journal publications. The July 2026 composition argument, the charmonium-mixing study and the radial-hybrid study are treated here as preprints. Sun and colleagues’ production critique is a 2022 journal paper.

For the experimental results themselves, return to the BESIII measurements report. A missing decay is useful evidence when its scope is stated precisely. Establishing that the dominant internal contribution is gluonic requires the wider comparison to hold up as well.