Fully-Charmed Tetraquarks: Structure Determination from di-J/ψ Spectra and Reproducible Analysis
This article examines how fully-charmed tetraquark states—the cc c̄c̄ system seen in the di-J/ψ final state—move from the question of "whether they exist" to "what they actually are." It explains why the di-J/ψ channel is a relatively clean probe of multi-quark structure, how symmetry constraints couple spin with orbital angular momentum, and how to go from signal extraction and invariant-mass spectra to global likelihood fits, significance tests with the look-elsewhere effect, and spin-parity analysis. It compares four common interpretations—compact tetraquark, molecular state, charmonium excitation, and paired-charmonium kinematic effects—and shows how mass relative to threshold, width, angular distributions, phase motion, and yield ratios discriminate among them. Finally, it maps the analysis onto a reproducible scientific-computing pipeline covering data tiering and versioning, job scheduling, observability, machine-learning assistance, and a reproducibility checklist.
Introduction: From “Whether It Exists” to “What It Actually Is”
In hadron physics, mesons are built from a quark and an antiquark, and baryons from three quarks—the most classic classification framework. Quantum chromodynamics (QCD), however, does not forbid combinations with more quarks, and exotic hadrons such as tetraquarks and pentaquarks have long been part of theoretical predictions. A fully-charmed tetraquark is composed of two charm quarks and two anticharm quarks—the cc c̄c̄ system. Because the charm quark has a moderate mass, the bottom and top quarks are too heavy, and light-quark systems tend to be buried in the hadronic background, the fully-charmed tetraquark is often regarded as one of the relatively clean laboratories for studying multi-quark structure.
At the Large Hadron Collider (LHC), the LHCb, ATLAS, and CMS experiments have reported resonant structures in the di-J/ψ final state in sequence. Among them, X(6900) was found earlier by LHCb and subsequently confirmed independently by ATLAS and CMS; structures such as X(6600) and X(7100) have also been reported in public measurements from ATLAS and CMS. The values here are all approximate masses in MeV. When reading further, you can compare the public papers, preprints, and released experimental data from the LHCb, ATLAS, and CMS collaborations on the di-J/ψ invariant-mass spectrum, and understand the differences among experiments in their trigger, reconstruction, and statistical methods.
Discovery is only the first step. The peak position, width, angular distribution, and yield together carry information about spin, parity, and internal structure. Following a route of “signal extraction → structure determination → model discrimination → engineering reproduction,” this article lays out an understandable, reusable, and auditable analysis framework.
I. Why Focus on the di-J/ψ Final State
1. J/ψ Is a Clean Charmonium Probe
A single J/ψ has a mass of about 3.097 GeV, and a system of two J/ψ mesons has a production threshold of about 6.19 GeV, often rounded to about 6.2 GeV. Candidate fully-charmed tetraquark states usually sit near or slightly above this threshold, so the di-J/ψ final state is an important channel for locating such structures.
The J/ψ decays to μ+μ− with a branching fraction of about 6%. Muons have strong penetrating power, high trigger efficiency, and good momentum resolution, making them well suited for reconstruction in high-event-rate collider environments. The di-J/ψ final state ultimately corresponds to four muons, so J/ψ candidates can be built from clear pairwise combinations and then combined into a four-muon invariant-mass spectrum. Because the final state contains no light-flavor hadrons, contamination from the light-quark hadronic background is relatively low, and the signal-to-noise ratio is often better than for many hadronic decay channels.
2. Symmetry Constraints of the di-J/ψ System Require Spin and Orbital Angular Momentum Together
The quantum numbers of the J/ψ are J^PC = 1^—. In a simplified picture, the parity and charge conjugation of a two-J/ψ system can be written in terms of the relative orbital angular momentum L, but a full determination cannot rely on L alone. The two J/ψ mesons are identical bosons, so the orbital angular momentum L and total spin S must be considered together, taking into account the symmetry requirements on the spatial and spin wave functions when the two particles are exchanged.
Therefore, it is not advisable to write the allowed quantum numbers as an absolute list of forbidden states. A more robust approach is to focus on testing hypotheses such as 0^++ and 2^++ under common assumptions of low relative orbital angular momentum or natural parity; meanwhile, odd-parity, vector, or other anomalous contributions should be treated as systematic uncertainties, alternative hypotheses, and cross-check items—not simply excluded before the fit. This approach preserves the physical constraints while avoiding the mistake of presenting a theoretical simplification as an experimental conclusion.
II. Signal Extraction: From Collision Data to Invariant-Mass Spectrum
1. Data Selection and Reconstruction Workflow
Candidate events for a di-J/ψ analysis typically contain four muons. A typical workflow is as follows:
| Stage | Core task | Example threshold or check | ||
|---|---|---|---|---|
| Online trigger | Select di-muon or four-muon candidates from high-rate collisions | Stable di-muon trigger paths, calibratable efficiency | ||
| Offline reconstruction | Measure muon momenta and reconstruct decay vertices | Reliable single-muon reconstruction quality | ||
| Particle identification | Distinguish muons, pions, and kaons | Reduce misidentification of light hadrons | ||
| J/ψ combination | Combine two muons into one J/ψ candidate | Example J/ψ mass window: about 3.0 to 3.2 GeV | ||
| Four-muon combination | Combine two J/ψ candidates into a di-J/ψ system | Four-muon vertex points to a common location | ||
| Vertex quality check | Control accidental combinations | Example vertex-fit χ²/ndf: less than 3 to 5 | ||
| Kinematic selection | Suppress low-momentum background | Example muon pT: greater than 2 to 3 GeV | ||
| Pseudorapidity range | Ensure detector acceptance | Example | η | : less than 2.4 to 2.5 |
These values are only illustrative thresholds that explain the analysis logic. A real analysis must be determined by the specific experimental detector, trigger conditions, data-taking period, and simulation calibration. What matters is not memorizing a particular number, but understanding that every selection step must come with a clear efficiency measurement, background estimate, and systematic-uncertainty evaluation.
2. Background Modeling
The main background in a di-J/ψ analysis is usually non-resonant di-J/ψ production. In a single hard scattering, mechanisms such as gluon fusion can directly produce two charmonium states, forming a continuous distribution. There are two common ways to handle the background:
- Describe the background shape with a full simulation sample, then fit it to the data;
- Fit the data sideband regions directly with polynomials, spline functions, or other parametric functions.
In practice, analyses often need to compare several background functions and fold the shape variation into the systematic uncertainties. In addition, bottom-hadron decay chains, misreconstruction, and partial-phase-space events may also contribute to the background and need to be evaluated through simulation, data-driven methods, or sideband checks.
3. Invariant Mass and Likelihood Function Illustration
The code below only illustrates the basic structure of invariant-mass calculation and the extended Poisson likelihood, and cannot be used directly for a formal physics fit. A real fit requires strictly normalized signal and background probability density functions and a professional statistical framework to handle resolution convolution, efficiency corrections, and covariance.
import mathdef invariant_mass(e1, px1, py1, pz1, e2, px2, py2, pz2):"""Combine the energy and momentum of two leptons into an invariant mass, in GeV"""e = e1 + e2px, py, pz = px1 + px2, py1 + py2, pz1 + pz2return math.sqrt(max(e * e - px * px - py * py - pz * pz, 0.0))def breit_wigner(m, m0, gamma):"""Illustrative relativistic Breit-Wigner shape; normalization is required for real use"""return 1.0 / ((m * m - m0 * m0) ** 2 + (m0 * gamma) ** 2)def linear_bkg(x, x0, b0, b1):"""Illustrative linear background"""return max(b0 + b1 * (x - x0), 0.0)def extended_nll(bins, counts, params, signal_yield, frac_sig, b0, b1, x0=6.2):"""Illustrative only: params=[m0, gamma]; signal_yield is the overall event scale;frac_sig is the signal fraction; b0, b1 are the linear background parameters.A real fit must ensure the signal and background shapes are normalized and mustcombine resolution and efficiency corrections."""dx = bins[1] - bins[0]total = 0.0m0, gamma = params[0], params[1]for x, k in zip(bins, counts):shape = frac_sig * breit_wigner(x, m0, gamma)shape += (1.0 - frac_sig) * linear_bkg(x, x0, b0, b1)mu = signal_yield * shape * dxmu = max(mu, 1e-300)total += mu - k * math.log(mu)return total
Here signal_yield represents the overall event scale, frac_sig the signal fraction, and b0 and b1 control the linear background trend. A formal analysis also needs to write resolution effects as a convolution term, express efficiency as a function of mass or kinematics, and fold all parameter constraints and covariances into the fit.
III. Structure Determination: Likelihood, Significance, and Spin-Parity
1. Global Fit of Multiple Resonances
X(6600) and X(6900) differ in mass by about 300 MeV. If fitted individually, the high-mass and low-mass tails of the two peaks may overlap, causing parameter absorption or bias. A more robust approach is a global fit:
- The invariant-mass axis spans roughly 6.2 to 7.4 GeV;
- Multiple resonances are described with Breit-Wigner functions, added coherently or incoherently according to the physical hypothesis;
- The overall signal can be multiplied by a smooth phase-space factor;
- The detector mass resolution is handled by convolving with a Gaussian or other response function;
- The background shape participates in the fit simultaneously, avoiding the loss of statistical information from subtracting the background before fitting.
The key to a global fit is parameter interpretability. If the mass, width, yield, and phase of each resonance fail to converge stably, you need to check whether the initialization, background model, resolution model, and physical assumptions are overly complex.
2. Significance Testing and the Look-Elsewhere Effect
Whether a peak is significant can be judged with a profile-likelihood-ratio test statistic:
q = -2 ln(L0 / L1)
where L0 is the background-only hypothesis and L1 is the hypothesis including the target resonance. The larger q is, the more the data disfavor the background-only hypothesis.
However, if the analyst scans multiple windows across the mass spectrum, an additional-tries effect—the look-elsewhere effect—can arise. Some local fluctuations may appear significant simply because of the number of scans. A robust approach is:
- Define the scan range explicitly, for example from 6.2 to 7.4 GeV;
- Generate a large number of background-only samples with pseudo-experiments;
- Repeat the same scan and test procedure for each pseudo-experiment;
- Obtain the global p-value, then convert it to a global significance;
- When reporting, give both the nominal (local) significance and the global significance.
The number of pseudo-experiments is usually at least a thousand, and complex scans can reach several thousand or tens of thousands. For preliminary conclusions, it is appropriate to state “suggestive evidence”; only when the global significance is high enough should stronger wording such as “observed” or “discovered” be used.
3. Spin-Parity Analysis
A mass spectrum can locate a peak, but determining spin and parity usually requires angular-distribution information. In the rest frame of the resonance or an appropriately defined reference frame, the angular distribution of the final-state muons carries angular-momentum information. Common reasoning is as follows:
- A scalar 0^++ hypothesis often gives a relatively flat angular distribution;
- A tensor 2^++ hypothesis may show measurable anisotropy;
- Higher-spin or higher-orbital-angular-momentum hypotheses are usually suppressed by phase space;
- Different hypotheses can be written as partial-wave amplitudes, then compared via the likelihood.
This step is very sensitive to model dependence and systematic uncertainties. A real analysis must simultaneously check how muon reconstruction efficiency varies with the angular variables, the detector acceptance, the background angular distribution, the definition of the reference frame, and statistical fluctuations. Spin-parity conclusions are best cross-validated with multiple decay channels, multiple reference frames, and multiple fitting schemes.
4. Systematic Uncertainties and Verifiable Conditions
| Source | Possible impact | Common treatment |
|---|---|---|
| Energy scale and resolution | Peak-position and width shifts | Calibrate with known hadrons |
| Background shape | Signal-yield bias | Compare alternative functions |
| Spin-parity assumption | Yield and parameter shifts | Include alternative hypotheses and likelihood factors |
| Production mechanism | Model-dependent shape | Vary partial-wave assumptions |
| Detector efficiency | Distorted angular distribution and yield | Correct with simulation and data-driven methods |
| Statistical method | Significance-estimation bias | Pseudo-experiments and global significance tests |
Credibility comes not only from the fit results but from whether the results can be reproduced, whether they can be tested against alternative hypotheses, and whether the sources of uncertainty are clearly stated.
IV. Discriminating Among the Four Candidate Models
Interpretations of the fully-charmed tetraquark commonly include a compact tetraquark, a molecular state, a charmonium excitation, and kinematic effects from paired charmonium production. Not all of these can be judged by mass alone; they require combining mass, width, angular distribution, phase motion, and yield.
| Model | Typical expectation | Key discriminant |
|---|---|---|
| Compact tetraquark | Color-anti-symmetric diquark pairs recoupled into a color singlet | Mass above threshold, width, coupling strength |
| Molecular state | Nuclear-like binding of two charmonia | Proximity to threshold, binding energy, angular distribution |
| Charmonium excitation | Higher radial or orbital excitation | Other decay channels, spectral relations, width |
| Paired-charmonium kinematic effect | Triangle singularities near threshold, etc. | Phase motion, non-resonant features, angular distribution |
Take X(6900) as an example: its mass is about 6.9 GeV, while the di-J/ψ threshold is about 6.19 GeV, so it sits roughly 700 MeV above threshold. Mass measurements usually also carry an uncertainty of a few to tens of MeV, depending on the experimental statistics, energy-scale calibration, and fit model. For a simple molecular-state picture, staying close to the production threshold is a common expectation; if a structure sits clearly above threshold, a purely molecular interpretation faces the problem of too large a binding energy or too strong a coupling. But this judgment depends on the specific model and phase-space treatment, and a molecular state cannot be ruled out merely because the mass is “above threshold.”
The core handles for discrimination include:
- Mass relative to threshold: how far the structure sits from the di-J/ψ threshold;
- Width: a narrow state is closer to a bound structure, while a broad state may include strong coupling or kinematic effects;
- Angular distribution: different spin-parity and production mechanisms give different angular structures;
- Phase motion: a genuine resonance usually shows a describable phase variation;
- Yield ratios: different models can predict cross sections and branching fractions that differ by several times.
V. Migrating the Method to Reproducible Computing Engineering
A complete resonance analysis is essentially a scientific-computing pipeline: raw data pass through trigger selection, reconstruction, background modeling, fitting, significance testing, and model comparison, and every step should be reproducible, auditable, and traceable. The general capabilities of a cloud platform can carry this workflow, but security, access control, data governance, and cost control must be considered at the same time.
1. Data Tiering and Version Management
Raw collision data are large in volume and accessed infrequently, so they can be placed in a low-frequency access tier of object storage; reconstructed intermediate products go into the standard tier; final fit inputs are small and accessed frequently, so they can be cached in high-performance storage. Data versions must be fixed to avoid different training, fitting, or validation tasks using inputs of different versions.
2. Job Scheduling and Parallelization
Monte Carlo pseudo-experiments, alternative background samples, and partial-wave fitting tasks are usually independent of one another and well suited to parallel computing. A message queue can decouple task generation from task execution, and the container platform can auto-scale based on queue backlog. Pay attention to resource quotas and cost caps to prevent large numbers of invalid tasks from piling up due to erroneous parameters.
3. Full-Chain Observability
Logging services should record trigger efficiency, reconstruction quality, fit convergence status, likelihood values, and pseudo-experiment results. Monitoring and alerts should cover at least queue backlog, job failure rate, storage throughput, compute-resource utilization, and abnormal exits. When a fit fails to converge or samples are missing, it should be possible to quickly locate the specific data block, job ID, and parameter version.
4. Machine-Learning Assistance
Gradient-boosting trees, neural networks, and other methods can be used in trigger selection, background suppression, and event classification. But they must follow the same principles as the physics analysis: the training, validation, and test sets are kept strictly independent; feature selection must not introduce future information; offline metrics cannot replace online stability evaluation; and model interpretability and failure cases must be recorded.
5. Reproducibility Checklist
[ ] Code version, container image, and dependency library versions pinned[ ] Input data version and data lineage fully recorded[ ] Train/validation/test sets strictly split by random seed or time[ ] A baseline model or background-only fit kept as a control[ ] Random seeds, hardware environment, and parallel parameters recorded[ ] Key conclusions re-run at least twice[ ] Access control follows the principle of least privilege, with keys and credentials centrally managed[ ] Storage has versioning, backup, and lifecycle policies enabled[ ] Cost caps, budget alerts, and resource quotas configured[ ] Failed jobs support bounded retries, with failure reasons recorded[ ] A clear deletion policy defined for temporary data, logs, and caches
VI. Conclusion and FAQ
Research on fully-charmed tetraquarks pushes the question from “whether multi-quark states exist” to “what they actually are.” From the di-J/ψ invariant-mass spectrum, to the global likelihood fit, to spin-parity analysis in the angular distributions, every step depends on rigorous data governance and computing engineering. Only when the fit precision, significance, and model-discrimination power are all sufficient can exotic hadrons move from naming to structure.
FAQ
What is a fully-charmed tetraquark?
It is a four-quark system composed of two charm quarks and two anticharm quarks, with common final states including the di-J/ψ.Why is the di-J/ψ final state commonly used?
Because the J/ψ decays cleanly to a muon pair with good mass resolution and relatively low final-state background, making it well suited for reconstruction and measurement.How do you distinguish a compact tetraquark from a molecular state?
You cannot rely on mass alone; you need to combine distance from threshold, width, angular distribution, phase motion, and yield ratios. A molecular state depends more on near-threshold behavior, while a compact tetraquark requires checking the color structure and coupling strength.How is spin-parity determined?
Mainly through the final-state angular distributions and partial-wave amplitude fits, comparing the likelihoods of different J^PC hypotheses while checking detector efficiency and systematic uncertainties.What are common misjudgments?
These include treating a statistical fluctuation as a signal, ignoring the look-elsewhere effect, choosing the background shape improperly, failing to convolve the resolution correctly, not correcting the angular-distribution efficiency, and over-relying on a single fit model.
