Octet-Aware Quarkonium Evolution in a Bulk-Viscous Plasma

Heavy quarkonia, bound states of a charm or bottom quark with its antiquark such as the $J/\psi$ or $\Upsilon$, are used as thermometers of the quark-gluon plasma made in heavy-ion collisions. The more weakly bound excited states dissolve at lower temperatures than the ground states. Current models of this dissolution, including recent ones that let the plasma be out of local thermal equilibrium, assume the quarkonium is already a fully formed, color-neutral bound state that sits in a static screened potential. That assumption ignores how the pair is actually produced. The standard production theory, NRQCD factorization (non-relativistic QCD, an effective theory that separates the short-distance creation of the quark pair from its slow binding into a meson), says that many pairs are created carrying net color, in a "color-octet" state, and only later shed that color by emitting soft gluons. This proposal would evolve the full color-singlet plus color-octet quantum state of the pair through a plasma whose departure from equilibrium changes with time. The starting state is fixed by NRQCD, not by assuming a pre-formed singlet meson.

The practical payoff is concrete. The closest existing calculation explicitly fails to predict the suppression of excited states at high temperature, because their spectral peaks become too broad to fit as isolated resonances (Spectral functions of heavy quarkonia in a bulk-viscous quark gluon plasma). An open-quantum-system evolution, which tracks a density matrix and never fits a peak, does not have this failure mode. It also lets us ask a question no one in the retrieved literature has posed: is the ratio of excited-state to ground-state yields more sensitive to the plasma's viscous non-equilibrium, or to the poorly known octet fraction at production?

Core Claim

We claim that an evolution equation built from an NRQCD-motivated singlet/octet initial density matrix, with in-medium couplings derived from bulk-viscous distribution functions whose strength follows the expansion in time, can do two things. It can predict excited-state survival where static spectral-function methods break down. It can also show whether non-equilibrium effects and color-octet production can be told apart in the $\psi(2S)/J/\psi$ ratio. The direction of thermal non-equilibrium supplies the medium, through the deformed distribution $f = f_0 + \delta f$ with bulk-viscous strength $\Phi$. NRQCD supplies the initial condition and the color structure. The suppression observables measured in heavy-ion collisions supply the evaluation setting.

A rough scale estimate shows why the color structure cannot be ignored. Binding happens at the ultrasoft energy $m_Q v^2$. With $m_c \approx 1.28$ GeV and $v^2 \approx 0.3$ for charm (Heavy Quarkonium Physics), this energy is about 0.4 GeV, so the formation time is about 0.5 fm/$c$. Bottom gives a similar value ($4.22 \times 0.1$ GeV). The pair is therefore still in its octet-to-singlet transition while the medium is forming and expanding fastest. Treating it as a finished singlet meson from the start is an assumption, not a controlled approximation.

Relation to Prior Work

Three papers by Thakur, Hirono and Haque motivate the proposal and supply its medium model. They deform thermal quark and gluon distributions with a bulk-viscous correction controlled by $\Phi$ and compute the color dielectric function in the hard-thermal-loop approximation, a standard weak-coupling treatment of how soft fields are screened by the plasma. From this they build a complex Cornell potential, whose real part describes screening and whose imaginary part describes collisional damping.

The papers then use this potential in different ways:

All three treat $\Phi$ as a constant and the quarkonium as a pre-formed singlet. We reuse their hard-thermal-loop potential construction directly, but change two things. First, the potential becomes the Hamiltonian and dissipator of a time-dependent evolution instead of a static eigenvalue problem. Second, the state being evolved carries color-octet components.

The NRQCD side constrains the design. Inclusive J/psi photoproduction (Cacciari and Krämer) shows that octet matrix elements fitted at the Tevatron overpredict HERA photoproduction. The octet fraction at production is therefore not known to better than a large factor, and we must scan it rather than fix it. The double-$J/\psi$ calculations of He et al. show the opposite failure. At NLO, color-singlet-only predictions undershoot ATLAS and CMS by factors of 10 to 40 (He et al. 2025). Near threshold, the normalization depends strongly on how relativistic corrections are handled: a naive LO prediction gives about 17.43 nb, the corrected one gives about 12.02 nb, and LHCb measures about 7.9 nb at 13 TeV (He et al. 2024). We take two lessons from these results. Octet components cannot be set to zero, and absolute proton-proton normalizations carry uncertainties of order a factor of two. That is why we target ratios such as $\psi(2S)/J/\psi$ and $R_{AA}$ rather than absolute yields.

The non-equilibrium field-theory work of Cooper and collaborators on chiral condensates in an expanding plasma supplies a methodological warning. Its effective model is far from heavy quarks, but Quantum evolution of disoriented chiral condensates shows that an instantaneous quench, a sudden switch from vacuum to medium conditions, badly overestimates instability growth compared with continuous evolution in an expanding medium. We take this as a reason to evolve continuously rather than project the vacuum wave function onto in-medium states.

The retrieved file contains no open-quantum-system treatment of quarkonium. Equilibrium versions of such master equations likely exist outside what was retrieved. Our novelty claim therefore rests on the combination of three ingredients: time-dependent non-equilibrium transport coefficients, an NRQCD octet initial state, and the excited-state regime where Breit-Wigner extraction fails. It does not rest on the use of a master equation itself.

Research Program

The core object is the reduced density matrix $\rho(\tau)$ of the heavy pair, in relative coordinate and color (singlet or octet). It evolves in proper time $\tau$ by a master equation of Lindblad form, the standard form for a quantum system losing coherence to an environment:

$$\frac{d\rho}{d\tau} = -i[H(\tau),\rho] + \sum_k \left( L_k \rho L_k^\dagger - \tfrac{1}{2}\{L_k^\dagger L_k,\rho\} \right).$$

The pieces of this equation are defined as follows:

  • The Hamiltonian $H(\tau)$ contains the kinetic term and the real part of the singlet and octet in-medium potentials.
  • The jump operators $L_k$ encode color-changing ($1\leftrightarrow 8$) and momentum-kicking collisions. Their strength is fixed by the imaginary part of the potential and its color decomposition.
  • All medium inputs are computed from the bulk-viscous hard-thermal-loop dielectric function of Thakur and Hirono, evaluated at the local temperature $T(\tau)$ and viscous strength $\Phi(\tau)$.

The first build step extends that hard-thermal-loop construction from the singlet potential to the octet channel and to the color-changing correlator that sets the $1\leftrightarrow 8$ rates. The second step replaces constant $\Phi$ with a time-dependent $\Phi(\tau)$. We take this from first-order bulk-viscous Bjorken expansion (boost-invariant longitudinal flow), in which the bulk pressure scales with $\zeta/\tau$ ($\zeta$ the bulk viscosity). The ratio $\zeta/s$ to entropy density is allowed to peak near the pseudo-critical temperature $T_{pc} = 158$ MeV (Kumar et al.). The third step builds the initial density matrix. Its color-octet weight $w_8$ in each spin-orbital channel is set from NRQCD short-distance coefficients times long-distance matrix elements. Because of the universality crisis, $w_8$ is scanned between a singlet-dominated and an octet-dominated end point. The fourth step evolves $\rho$ to freeze-out and projects onto vacuum $J/\psi$, $\psi(2S)$, $\Upsilon(1S)$ and $\Upsilon(2S)$ wave functions to obtain survival probabilities, $R_{AA}$, and yield ratios.

Baselines

The naive reference point is local thermal equilibrium with no viscous deformation. It sets $\Phi = 0$ at all times and starts from a pure singlet, so it reduces our machinery to an equilibrium singlet master equation. Any effect we attribute to non-equilibrium or to octet production must show up as a difference from this run. Both runs use the same temperature profile, the same projection, and the same numerical grid.

The strongest existing competitor is the static spectral-function approach of Thakur and Hirono. It uses constant $\Phi$, solves the Schrödinger equation in the complex potential, and extracts widths by Breit-Wigner fits as in their 2022 paper. It is the only method in the file that has been confronted with the CMS $\psi'/J/\psi$ ratio under a non-equilibrium medium. For a fair comparison we rerun it with their published potential and parameters. We feed it the same temperature history and the same scan over $\Phi$ values. The file reports no numerical value for their CMS comparison or for their $R_{AA}$: no reported result found on Lacuna for the $\psi'/J/\psi$ value or $R_{AA}$ numbers in that work. We will therefore reproduce their curves ourselves before comparing.

Beating this baseline means three things. Our method must match it where it is valid: ground states, and excited states at temperatures where the widths are narrow. It must also return finite, stable $\psi(2S)$ and $\Upsilon(2S)$ survival probabilities in the high-temperature region where their fits fail. Finally, its $\psi(2S)/J/\psi$ prediction must fall closer to CMS than theirs at matched parameters, or with a smaller parameter-driven spread.

Two ablations isolate the key ingredients. The first sets $w_8 = 0$ (singlet-only initial state) while keeping time-dependent $\Phi$. The second keeps the octet initial state while freezing $\Phi$ at its time-averaged value. Each uses the identical grid and temperature profile as the full model.

Why not just…?

Why not just use equilibrium hydrodynamics with screened potentials? This is the $\Phi = 0$ baseline above, so it is tested directly rather than dismissed. Thakur and Hirono found that ground-state $R_{AA}$ is largely insensitive to $\Phi$. For $J/\psi$ and $\Upsilon(1S)$ the equilibrium model may therefore be adequate, and we expect to report that. The non-equilibrium medium has to earn its place in the excited states and near $T_{pc}$, where those authors predict non-monotonic behavior. If our excited-state results also turn out insensitive to $\Phi(\tau)$ once evolution is continuous, that is a reportable negative result.

Why not just evolve a color singlet and ignore octets? At high transverse momentum the singlet-only picture is empirically wrong in vacuum, undershooting double-$J/\psi$ data by factors of 10 to 40 (He et al.). Whether that matters in the plasma is an open question. Color-changing collisions might scramble singlet and octet so fast that the initial $w_8$ is forgotten. The $w_8 = 0$ ablation settles this. If the freeze-out ratios do not move across the $w_8$ scan, the octet ingredient is unnecessary for these observables and the simpler model wins.

Why not just use a classical rate equation with survival $\exp(-\int \Gamma\, d\tau)$, taking $\Gamma$ from the imaginary potential? This is the most serious threat. It is cheap, uses the same medium inputs, and may reproduce ground-state results. It differs from the master equation only through quantum coherence: the dissociated pair can re-form a bound state, and the states mix. These effects should matter most when the width is comparable to the binding energy, which is exactly the excited-state regime where the static method fails. We will run the rate equation with identical $\Gamma(\tau)$ and $w_8$. If it agrees with the master equation for $\psi(2S)$ and $\Upsilon(2S)$ within the parameter-scan spread, we will adopt it and say so.

Why not just use an instantaneous quench? In the chiral-condensate setting, Cooper et al. showed that a sudden quench exaggerates non-equilibrium effects relative to continuous evolution. We expect a similar bias here, toward too much early dissociation. A projection run with the sudden switch and the same final medium will quantify that bias.

Proposed Experiments and Decisive Checks

The first check is validation. In the limit of a static medium, constant $\Phi$, singlet initial state, and narrow widths, the long-time master-equation spectrum must reproduce the masses, binding energies, and widths of Thakur and Hirono at $T = 0.2$ GeV across their $\Phi$ values. It must also reproduce the lowered melting temperatures of Thakur, Hirono and Haque.

The second check is extension. We compute $\psi(2S)$ and $\Upsilon(2S)$ $R_{AA}$ across the temperature range where the Breit-Wigner extraction failed, and confirm numerical convergence in grid size and time step.

The third check, the central result, is a discrimination test. Over a two-dimensional scan in peak $\zeta/s$ and $w_8$, we map the freeze-out $\psi(2S)/J/\psi$ and $\Upsilon(2S)/\Upsilon(1S)$ ratios and ask whether the two parameters move the ratios along distinguishable directions. If they do, quarkonium ratios can constrain plasma non-equilibrium only once the octet fraction is fixed independently. If they are degenerate, the existing literature's attribution of ratio changes to bulk viscosity is not identifiable from these data alone. Either answer is useful.

The fourth check compares against the CMS $\psi'/J/\psi$ measurement used by Thakur and Hirono. We will take the data points from the CMS publication itself, because the file records no numerical value.

The final check is the head-to-head against the classical rate equation and the quench, which decides whether the quantum, octet-aware machinery is needed at all.