Quarkonium correlators and spectral functions at zero and finite temperature

The thing worth your time here is the correlator analysis, not the spectral functions. The authors run five lattice spacings with two anisotropies in the Fermilab formulation and compare each finite-temperature correlator to one reconstructed from the zero-temperature spectral function through the finite-T kernel. That ratio, G/G_recon, is the load-bearing evidence, and it is a genuine comparison against an independently determined baseline rather than a circular one. The pseudoscalar ratio stays near unity to about 1.5 Tc on the two finest lattices; the scalar, axial and tensor ratios rise by factors of two to four just above Tc. That contrast is the real result, and it is robust in the sense that it does not depend on the MEM inversion at all. The second contribution is the MEM reformulation in Section III, which turns the maximization into a minimization in data-dimension space and, by the authors' account, converges where Bryan's algorithm stalls on long charmonium correlators. That looks like a genuine practical improvement, and the comparison in the appendix backs it up. The systematic-error program is also real: matched tau_max and N_data baselines, jackknife errors on binned spectral functions, a finite-volume check, and an honest appendix comparing MEM masses and amplitudes against two-exponential fits. Credit where due, this is careful work and the authors are candid about their own weak points, including the default-model dependence of the reconstruction and the ad-hoc renormalization constants in Table VII. Now the soft spot, and it is the abstract. [abstract:s3] states 1P dissolution just above Tc as a finding. The body does not support that strength. Section VI.B says the enhancement 'indicates strong modification ... and, possibly, the dissolution' of 1P states, and it also reports that the enhancement is largest on the coarsest lattice and shrinks toward finer spacing. A signal that decreases with the cutoff cannot ground a flat claim. Worse, the paper itself raises the alternative mechanism in Section X for bottomonium P-states, citing Ref. [11]: the change may come from modification of the continuum part rather than ground-state melting. That mechanism is channel-generic and nothing in the charmonium analysis excludes it. The finite-T P-wave spectral functions cannot settle the question either, since P[alpha|D_m] has multiple maxima for most default models. So the stress-test note is right, and I would push it further: the abstract should be rewritten to match the body. The reader's MEM-prior worry is also fair but less damaging, because the correlator ratios carry the 1S claim independently. Who should read this: lattice people working on quarkonium spectral functions, and anyone who wants the correlator-level statement rather than the spectral-function one. It deserves referees, mainly to force the abstract into line with Section VI.B and to ask for a cleaner statement of what the P-wave enhancement can and cannot distinguish.

The abstract (abstract:s3) asserts 'dissolution of 1P quarkonium states right above the deconfinement temperature' as a finding. The body does not support that strength. Section VI.B states the scalar/axial/tensor enhancement 'indicates strong modification ... and, possibly the dissolution of 1P charmonia states' — 'possibly', not established. Three concrete problems sit under the flat abstract claim. (1) The enhancement is cutoff-dependent: Section VI.B says 'the magnitude of the enhancement is largest on the coarsest lattice and decreases with decreasing lattice spacing', and 'the cutoff dependence of G/G_recon is larger in the scalar channel than in the pseudo-scalar one'. A signal that shrinks toward the continuum limit cannot ground a flat dissolution claim. (2) The correlator ratio G/G_recon is an integral constraint, not a peak-existence test: a spectral function whose 1P peak has broadened but whose low-omega weight is redistributed can reproduce the same ratio. The paper itself raises exactly this alternative in Section X for bottomonium P-states, citing Ref. [11]: the large change 'may be due to the modification of the continuum part of the spectral functions and not to the dissolution or strong modification of the ground state'. That mechanism is channel-generic and is not excluded for charmonium. (3) The finite-T P-wave spectral functions are uncontrolled: Section VII.B reports multiple maxima of P[alpha|D_m] for most default models, so the MEM reconstruction cannot adjudicate. The reader's MEM-prior concern is real, but for the 1P half of the abstract the more load-bearing gap is that the correlator evidence is ambiguous between ground-state melting and continuum modification, and the paper's own text says so.

The reader's verdict (CONDITIONAL) and its weakest-assumption framing (MEM prior dependence) are sound and I do not overturn them. My concern is adjacent but distinct and, for the abstract's second clause, more load-bearing. The reader treats the correlator ratios as 'the most robust part of the paper because they do not depend on the MEM inversion', and for the 1S survival claim that is fair: G/G_recon near unity up to ~1.5 Tc on the two finest lattices (Figs. 8-9) is a genuine, prior-independent statement, and the paper is appropriately hedged there ('small but statistically significant deviations' at 1.5 Tc, and the 2 Tc conclusion 'could be premature'). For the 1P dissolution claim, however, the correlator ratio is not decisive, because the same ratio rises whether the 1P peak melts or the continuum is modified. The paper says this itself in Section X, citing Ref. [11], and Section VI.B only claims 'possibly the dissolution'. The abstract then states dissolution flatly. Add the cutoff dependence the authors report (enhancement largest on the coarsest lattice, shrinking with finer spacing) and the uncontrolled P-wave MEM (multiple P[alpha|D_m] maxima, Section VII.B), and the 1P clause is the least secure part of the central claim. This is an internal-tension concern (abstract vs body), not a disagreement with consensus, and it routes through correctness risk. The paper deserves credit for its systematic-error program (five spacings, two anisotropies, matched-time-extent zero-T baselines, jackknife errors, finite-volume check) and for stating the continuum-modification caveat at all; the fix is to soften the abstract's 1P clause to match the body and to run the continuum-only test above. That would not change the CONDITIONAL verdict, so I recommend UNCHANGED.

The systematic-error program is the paper's real contribution and is unusually thorough for its time: five lattice spacings, two anisotropies, a finite-volume check at β = 6.1 (Fig. 11), jackknife errors on binned spectral functions (Eq. 32), and — most importantly — a zero-temperature baseline reconstructed with the same number of time slices and the same τ_max as each finite-temperature set (Fig. 16), which removes the dominant trivial source of apparent temperature dependence. The new MEM algorithm is a genuine, reusable methodological advance: recasting the problem as the minimization of U in Eq. (19) with the positivity of ∂²U/∂s_i∂s_j stated, and demonstrating convergence where Bryan's algorithm required truncating to t_max = 24 (Figs. 27, 28), is a concrete and checkable improvement that later work can adopt. The identification of a transport contribution in the vector correlator (Fig. 15) is a falsifiable, physically meaningful statement that connects to heavy-quark diffusion. The conclusions are appropriately hedged in the body and the paper is candid about what it cannot do (bottomonium finite-T spectral functions, P-wave MEM reconstruction). If the 1S survival result and

This paper asks a question that heavy-ion physics needs answered: when the quark-gluon plasma forms, which quarkonium bound states actually disappear, and at what temperature? The authors compute charmonium and bottomonium correlation functions on quenched anisotropic lattices at several lattice spacings, then invert those correlators into spectral functions with a Maximum Entropy Method built on a new minimization algorithm. Their diagnostic is a ratio: the measured finite-temperature correlator divided by the correlator one would get by feeding the zero-temperature spectral function back through the finite-temperature kernel. If that ratio stays at one, the spectral function has not changed. They report that the ratio stays near one in the pseudoscalar channel up to roughly $1.5T_c$, but rises by factors of two to four in the scalar, axial-vector and tensor channels already near $1.1$-$1.2T_c$. If that reading holds, the $1S$ states ($J/ψ$, $η_c$) are still recognizable bound states well inside the plasma while the $1P$ states ($χ_c$, $h_c$) are already gone.

What carries the argument. Two pieces of machinery carry the argument. The first is a reformulation of the Maximum Entropy Method: instead of searching a large space of spectral functions, the authors show that the stationarity condition can be rewritten exactly in terms of $N$ variables $s_i$, one per data point, giving $σ(ω)=m(ω)exp[∑_i s_i K(ω,τ_i)]$ and reducing the problem to minimizing a function $U(s)$ whose Hessian is positive definite. This makes the inversion as fast as a $χ^2$ fit and, the authors report, numerically stable enough to use all $N_t/2$ time slices where the Bryan algorithm converged only with truncated data. The second is the ratio $G(τ,T)/G_{recon}(τ,T)$, with $G_{recon}$ built from the zero-temperature spectral function and the finite-temperature kernel $K(ω,τ)=cosh(ω(τ-1/2T))/sinh(ω/2T)$. Dividing out the trivial kernel dependence isolates the change in the spectral function itself, which is a much more robust signal than the shape of a single reconstructed peak.

Core claim. The central claim is a two-part temperature hierarchy for quarkonium in a deconfined medium. Using the ratio $G/G_{recon}$ of the lattice correlator to the correlator reconstructed from the zero-temperature spectral function, the authors find the pseudoscalar channel essentially unchanged up to about $1.5T_c$ on their two finest lattices ($a_t^{-1}=8.18$ and $14.12$ GeV), with medium modifications turning on only above that; the same near-flatness is seen in the vector channel apart from an additional low-frequency transport contribution. In the scalar, axial-vector and tensor channels, which host the $1P$ states, the same ratio departs from one sharply above $T_c$ and grows with decreasing $τ$, which the authors read as strong modification or dissolution of the $1P$ charmonia at $T ≃ 1.1$-$1.2T_c$. A Maximum Entropy reconstruction of the scalar spectral function at $1.16T_c$ is shown to differ visibly from the zero-temperature one reconstructed with the same number of time slices, supporting the correlator-level conclusion. In the bottomonium sector the $1S$ correlators stay flat to much higher temperatures, while the $1P$ bottomonium correlators change even more strongly than their charmonium counterparts, which the authors flag as unexpected.

The premise it rests on. The load-bearing premise is that the reconstructed finite-temperature spectral function is fixed by the lattice data rather than by the entropy prior, yet the authors themselves report that at $1.2T_c$ the first peak is absent for some default models and that at $2T_c$ a different default model shows almost no temperature dependence at all.

Quarkonium suppression in heavy-ion collisions should be sequential in the way the paper describes: $1P$ charmonia melt almost immediately above $T_c$, while $1S$ charmonia need temperatures around $1.5T_c$ or above.

Because the $1S$ states survive, the low-energy part of the vector spectral function at finite temperature should contain a heavy-quark transport peak in addition to the $J/ψ$ peak, and the authors estimate the mean thermal velocity squared as roughly $0.13$ at $1.2T_c$ and $0.18$ at $1.5T_c$.

Cutoff effects in the temperature dependence of quarkonium correlators are modest: the authors argue the correlators can be studied reliably already at spatial lattice spacings around $a_s^{-1}=2$ GeV, which matters for extending the work to full QCD where the lattice spacing is the limiting resource.

The survival of $1S$ states is not established by the shape of the reconstructed peak alone, since that shape depends on the default model; the claim rests on the default-model-independent statement that the pseudoscalar correlator barely changes below $1.5T_c$.

The two diagnostics disagree in strength: the correlator ratio is a robust, prior-independent statement, while the spectral-function shape is not. A reader should treat the $1S$ survival claim as supported mainly by the ratio and the $1P$ dissolution claim as supported by both, since the $1P$ enhancement is large enough to be visible without any inversion at all.

The unexpected strength of the $1P$ bottomonium signal is a useful anomaly: if the effect were really about binding energy or state size, bottomonium $1P$ should sit between charmonium $1S$ and charmonium $1P$ in robustness, not beyond charmonium $1P$. That ordering suggests the effect lives in the high-frequency continuum of the spectral function rather than in the bound-state peak, which is a te

The transport-peak extraction, $-(G_V-G_V^{recon})/G_{00}$, is a template that could be pushed further: with higher statistics and several lattice spacings one could try to fit the curvature that a finite heavy-quark diffusion constant would imprint, turning a qualitative plateau into a numeric bound on $D$.

Because the whole study is quenched, the natural next check is whether dynamical sea quarks move the $1.5T_c$ survival threshold; the authors themselves note that no quantitative estimate of sea-quark effects exists yet.

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review