Scalar nonet quarkonia and the scalar glueball: mixing and decays in an effective chiral approach

This is a solid piece of 2005-era scalar spectroscopy. The setup is a three-state mass matrix — nn, glueball, ss — with a flavor-mixing parameter epsilon added on top of the standard glueball mixing, and the decay amplitudes are written out in full in Appendix B. That last point matters: the calculation is checkable, and the two-photon ratios in Eqs. 45, 54, 67, 70 are genuine parameter-free consequences of the fitted mixing matrix, not fit inputs. Those are the paper's most useful output. The authors also present two solutions honestly rather than picking a favorite, and they state the chiral-scale problem themselves in the introduction — tree-level ChPT at 1.5 GeV is a phenomenological tool, not a controlled expansion. Credit for saying so up front rather than burying it.

The central quantitative claim is that the mixing matrices in Eqs. 42 and 51 assign a definite glueball content to f0(1500) (solution I) or f0(1710) (solution II). Those matrices are not determined by the mass sector alone: three masses (1370, 1507, 1714 MeV) cannot fix five mass/mixing parameters (MN, MG, MS, f, ε). The decay ratios break the degeneracy, and every decay amplitude in Eqs. B7–B10 multiplies the mixing elements B_ij by the same two couplings csd, csm. So the mixing angles and the couplings are entangled in the fit. The fit's χ² is dominated by K*0→Kπ, which is underestimated by a factor ~5 (χ²=17.59 of 29.01 in Table 1; 18.48 of 24.61 in Table 4). The paper's own cross-check in §III.D is telling: fixing csd=17.94 MeV, csm=7.35 MeV to reproduce K*0→Kπ and then refitting the remaining parameters gives χ²~500. That is a factor-2 inconsistency in csd between the isoscalar and isodoublet sectors. Because B_ij multiplies csd, csm in every amplitude, a factor-2 ambiguity in the couplings can trade against the mixing angles. The paper reports refits with K*0 excluded (Eqs. 49 and 56) and asserts 'the discussion about the isoscalar states and a0(1450) remains unchanged,' but it does not quote the resulting B matrices for those fits. The qualitative claim that some mixing is required is reasonably safe (with cgd=cgm=0 and f=0 the glueball would be a stable pure state, excluded by the observed widths). What is not shown to be robust is the quantitative assignment of which resonance is glueball-dominant.

The reader's weakest_assumption — tree-level ChPT at 1.5 GeV — is a real limitation, but the authors state it themselves in the introduction ('a chiral approach cannot be rigorously justified at this energy scale'), so it is an acknowledged caveat rather than a new finding. I agree it caps confidence, but it is not the most specific unresolved issue. The more pointed concern is the coupling–mixing degeneracy: csd and csm multiply every B_ij in the decay amplitudes, the mass sector alone cannot fix f and ε, and the fit's χ² is dominated by a single point that the model misses by a factor of five. The paper's own §III.D cross-check (χ²~500 when csd, csm are pinned to K*0) shows the two sectors are mutually inconsistent at the factor-2 level, yet the resulting uncertainty on the glueball fractions is never quantified. This does not overturn the qualitative conclusion that mixing is required and direct glueball decay is not, but it does weaken the specific 'f0(1500) dominant' vs 'f0(1710) dominant' assignment that the reader's strongest_claim carries. Credit where due: the parameter-free two-photon ratios (Eqs. 45, 54, 67, 70) and the ηη′/ππ ratios are the paper's most independently checkable output and differ sharply between solutions, so they are a genuine falsification handle. The K*0 mismatch, the hand-imposed (Γ_f3)_2P = 140±10 MeV saturation constraint, and the self-cited cse normalization are all honestly flagged. CONDITIONAL remains the right verdict; my concern reinforces it rather than moving it to REJECT. I did not reproduce the numerical fit, so confidence stays moderate.

The paper is a careful and unusually transparent phenomenological study of a genuinely open problem. Its strengths are real and worth naming: the effective Lagrangian (Eq. 2), the mass matrix (Eq. 16) and the complete decay amplitudes (Eqs. B7-B20) are given in full, so the calculation is reproducible; the two-photon ratios of Eqs. (45), (54), (67) and (70) are genuine parameter-free consequences of the fitted mixing matrix and are not used as fit input, which makes them falsifiable predictions; and the paper does not hide its difficulties, presenting two competing solutions, a separate section (§III.D) on the K*0(1430) mismatch, and an explicit statement in §I that the chiral framework is not rigorously justified at this scale. The qualitative conclusion that some glueball-quarkonium mixing is needed is well supported. What is not yet established to the standard the central claim requires is the quantitative assignment of the gluonic component to a specific resonance.

Three neutral scalar resonances sit in the 1.3–1.7 GeV range — $f_0(1370)$, $f_0(1500)$ and $f_0(1710)$ — and nothing in QCD says outright which is quark-antiquark and which is the gluonic bound state that lattice calculations put in the same region. The paper's proposal is that all three are quantum mixtures of a non-strange quarkonium pair $N = (ubar u + dbar d)/sqrt 2$, a strange pair $S = sbar s$, and a flavor-blind glueball $G$, with the mixing generated by a $3 times 3$ mass matrix whose off-diagonal entries are two numbers: a glueball-quarkonium strength $f$ and a direct quarkonium flavor-mixing term $epsilon$. Diagonalizing that matrix and feeding the resulting decay amplitudes into twelve measured masses, widths and ratios from the 2004 Particle Data Group tables yields two acceptable solutions, which differ in which resonance carries most of the glue. What the paper argues is the robust part is the mixing itself: setting $f = 0$ does not describe the data, whereas switching on a separate direct glueball-to-two-pseudoscalar coupling buys almost nothing. If that reading holds, the scalar glueball is not a particle to be found but a component to be apportioned, and the pattern of $pi pi$, $K bar K$ and $eta eta$ branching ratios is the instrument that apportions it.

What carries the argument. The load-bearing object is the $3 times 3$ mass matrix in the basis $(N, G, S)$, with diagonal entries $M_N^2$, $M_G^2$, $M_S^2$ and off-diagonal entries built from two parameters: $sqrt 2 f$ connecting the glueball to $N$ and $f$ connecting it to $S$ (flavor blindness, so the glue couples to the singlet combination only), plus $epsilon$ connecting $N$ and $S$ directly, which the paper attributes to instanton effects and takes positive. Diagonalizing this matrix gives the three physical masses and the orthogonal matrix $B$ whose rows are the glueball and quarkonium contents of $f_0(1370)$, $f_0(1500)$ and $f_0(1710)$. The second piece of machinery is the decay Lagrangian: two quarkonium couplings $c_{sd}$ and $c_{sm}$ multiplying $langle S u_mu u^mu rangle$ and $langle S chi_+ rangle$, and two analogous glueball couplings $c_{gd}$ and $c_{gm}$, so that a physical resonance decays in proportion to its $B_{iN}$, $B_{iS}$ and $B_{iG}$ weights. Every strong width in the paper is a sum of three such weighted amplitudes, and the interference between them — constructive or destructive depending on the signs in $B$ — is what produces the predicted $K bar K / pi pi$ and $eta eta / pi pi$ ratios. The third piece is the two-photon coupling, where the quarkonium piece is fixed by a single constant $c_{se}$ and the glueball-photon coupling $c_{ge}$ is set to zero on the argument that gluons do not couple directly to photons.

Core claim. The central claim is that a tree-level chiral Lagrangian containing a scalar nonet plus an extra flavor-singlet glueball field, with flavor-blind glueball-quarkonium mixing and a possible direct $N$–$S$ mixing, can reproduce the accepted scalar meson data above 1 GeV with two distinct parameter sets. In solution I the fit gives $M_N = 1.455$ GeV, $M_G = 1.490$ GeV, $M_S = 1.697$ GeV, $f = 0.065$ GeV$^2$, $epsilon = 0.211$ GeV$^2$ with $chi^2 = 29.01$ for twelve observables; the mixing matrix is dominated by its diagonal, and $f_0(1500)$ carries the largest glueball fraction. In solution II the fit gives $M_N = 1.298$ GeV, $M_G = 1.513$ GeV, $M_S = 1.593$ GeV, $f = 0.400$ GeV$^2$, $epsilon = 0.015$ GeV$^2$ with $chi^2 = 24.61$; here the glueball is spread across the states and $f_0(1710)$ holds the largest gluonic component together with a large $sbar s$ piece. Both solutions share one conspicuous failure, which the authors state plainly: the predicted $K^{*0}(1430) to K pi$ width comes out near 60 MeV against a measured $273 pm 51$ MeV, and this single point dominates $chi^2$ in every fit. Allowing a direct glueball decay coupling $c_{gd}$ in the flavor-symmetric limit returns $|c_{gd}/c_{sd}|$ of order 0.1–0.2, i.e. no phenomenological demand for it, in line with large-$N_c$ counting and against at least one earlier fit that found direct glueball decay dominant. Predictions that go beyond the fitted set include two-photon width ratios, $Gamma_{f_1}:Gamma_{f_2}:Gamma_{f_3}:Gamma_{a_0^0} = 1:0.305:0.002:0.471$ in solution I, and the branching ratios $Gamma_{f_0(1710) to eta eta'}/Gamma_{f_0(1710) to pi pi}$, which solution I puts at 1.59 against an experimental upper bound of 0.18.

The premise it rests on. The whole quantitative result rests on treating a tree-level chiral Lagrangian as a usable numerical tool at 1.5 GeV, an energy where the authors themselves say loop corrections could be large and the expansion cannot be rigorously justified.

If the mixing picture is right, no physical scalar in the 1.3–1.7 GeV region is a pure glueball, so a search for 'the' scalar glueball as a single resonance is misposed; the measurable is the glueball fraction of each state, which differs sharply between the two solutions.

The two-photon partial widths of all three isoscalars plus $a_0^0(1450)$ are fixed in ratio independently of the unknown quarkonium-photon coupling, giving $1 : 0.305 : 0.002 : 0.471$ in solution I and $1 : 0.253 : 0.055 : 0.493$ in solution II — a target for two-photon collider analyses.

Measured violations of the Gell-Mann–Okubo relation in the scalar octet come out small ($gamma_{S8} = 0.225$ and $0.139$ in the two solutions), which supports continuing to use chiral and large-$N_c$ bookkeeping for the scalar nonet even at 1.5 GeV.

The ratio $Gamma_{f_0(1710) to eta eta'}/Gamma_{f_0(1710) to pi pi}$ is predicted to be well above the current experimental bound in solution I and near it in solution II, so that single number separates the two scenarios.

The persistent factor-of-five underestimate of $Gamma_{K^{*0}(1430) to K pi}$ is not repaired by any variant tried, including freeing the glueball couplings, which the authors take as evidence that the two-body tree-level description of the isodoublet is incomplete rather than that the mixing scheme is wrong.

The two solutions are distinguished mainly by where the glue sits, so an independent determination of the glueball fraction of $f_0(1710)$ — from radiative or two-photon couplings, or from production rates in gluon-rich channels — would collapse the ambiguity faster than further branching-ratio fits.

Because the direct glueball decay coupling is driven to near zero while the mixing is not, the natural reading is that glue reaches pseudoscalar pairs only after converting into quarkonium; that suggests glueball production and decay cross sections should track the quarkonium content of the state rather than its gluonic content.

The $K^{*0}(1430)$ shortfall and the $f_0(1710) to eta eta'$ excess both point the same direction: the two-body tree amplitudes are missing a channel. Coupling the scalars to the broad low-lying states $sigma$ and $K^{*0}(800)$, as the authors suggest, would be the obvious next term, and it would also shift the fitted mixing angles.

Since the flavor-mixing parameter $epsilon$ is large in solution I and negligible in solution II, the instanton-induced $N$–$S$ mixing that is well established in the pseudoscalar nonet is, on this evidence, not established in the scalar nonet; a lattice determination of the sign and size of that off-diagonal entry would decide it.

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