WorksPhysics and Astronomy
NNLO QCD corrections to radiative and hadronic decays
Abstract
We present the complete next-to-next-to-leading-order (NNLO) QCD corrections to radiative and inclusive hadronic decays in the color-singlet framework at leading order in the heavy-quark velocity expansion. For radiative decay, the NNLO correction reshapes the photon energy spectrum and lowers the partial width in the fiducial window , bringing the central prediction into good agreement with reconstructed CLEO data. For inclusive hadronic decay, the NNLO correction provides a large negative shift that lowers the predicted width from its inflated NLO value to a level fully consistent with the experimental determination. Although both individual decay widths exhibit larger renormalization-scale uncertainties at NNLO, their ratio demonstrates a significantly reduced scale dependence due to the cancellation of common short-distance dynamics and the nonperturbative matrix element. Using the window ratio inferred from CLEO measurements, we extract the strong coupling constant , compatible with the world average, and evaluate the sensitivity to uncalculated higher orders via the unexpanded quotient.
Introduction
Annihilation decays of heavy quarkonia provide clean laboratories for probing quantum chromodynamics (QCD) at intermediate energy scales [2, 3]. Within nonrelativistic QCD (NRQCD), decay widths factorize into short-distance coefficients, calculable as perturbative expansions in the strong coupling , and nonperturbative long-distance matrix elements (LDMEs) organized by powers of the heavy-quark velocity [4, 5]. Bottomonium () is especially well suited for such precision studies: the bottom-quark mass GeV lies well within the perturbative domain, and relativistic corrections () are strongly suppressed compared to charmonium.
The dominant annihilation modes of the state are inclusive hadronic decay () and radiative decay (). Next-to-leading-order (NLO) QCD corrections to both channels were established decades ago [6, 7], but revealed persistent theoretical difficulties. In the hadronic channel, the large positive NLO correction increased the predicted width well beyond the total measured width [6, 8], posing a longstanding puzzle for the perturbative convergence of vector quarkonium annihilation. In the radiative channel, the NLO correction modified the photon energy spectrum, yet experimental measurements cover a restricted kinematic window, requiring a reliable theoretical understanding of the spectral shape to interpret integrated yields [9] [1]. Subsequent findings of large NNLO effects in pseudoscalar quarkonia [10] and [11] strongly motivated evaluating both annihilation channels at next-to-next-to-leading order [12].
In this Letter, we compute the complete NNLO QCD corrections to radiative and inclusive hadronic decays at leading order in the velocity expansion. We determine the full differential photon spectrum—incorporating collinear photon fragmentation—and the partial width within the CLEO fiducial window, alongside the inclusive hadronic width. The NNLO corrections substantially decrease both decay rates, bringing the photon spectrum into close alignment with CLEO data and resolving the longstanding overprediction of the hadronic width. Furthermore, the ratio of the two widths eliminates the leading color-singlet LDME and exhibits markedly reduced scale sensitivity, providing a clean determination of from bottomonium decays.
Calculation.—At leading order in the heavy-quark velocity expansion, the bottom-quark mass is fixed by kinematics to , with GeV. In NRQCD factorization, the differential decay widths for radiative and hadronic decays are expressed in a unified form:
where denotes the integrated short-distance coefficient. The color-singlet LDME is defined at the NRQCD factorization scale and is related to the radial wave function at origin by .
The computation is performed using conventional dimensional regularization with four massless light flavors () and an on-shell heavy bottom quark. The strong coupling and the NRQCD composite operator are renormalized in the scheme. Multi-loop and multi-leg amplitudes are generated and simplified algebraically with FeynArts and FORM [13, 14]. The resulting Feynman integrals are reduced to master integrals via integration-by-parts identities using Kira 3.1 [15, 16, 17] and SecDec [18], and are evaluated with high numerical precision using AMFlow [19]. Infrared divergences in the real-emission phase space are isolated and subtracted using local analytic sector subtraction [20, 21, 22] and nested soft-collinear subtraction [23, 24]. After UV renormalization, the QCD infrared poles cancel analytically upon combining the real and virtual corrections with the NRQCD matching contributions [25, 26]. For phase-space integration, the interference between the two-loop virtual and tree-level amplitudes is integrated using Gauss-Kronrod (GK) quadrature, whereas the remaining phase-space contributions are evaluated via Monte Carlo integration. A comprehensive account of technical details and analytic structures will be reported in a forthcoming publication.
For the central predictions, we employ three-loop running with [3]. The wave function at the origin is taken as [11, 27], and the electromagnetic coupling is [28]. The renormalization scale is varied from to around the central value , with and all other inputs held fixed.
In terms of , the short-distance coefficients expand as
Renormalization group invariance and NRQCD matching determine their scale dependence [29, 30]:
where barred quantities correspond to and , , , , , and the power is 2 for radiative decay and 3 for hadronic decay.
Radiative decay.—In the rest frame of the , we define the photon energy fraction as . At NNLO, the double-real radiation channel includes photon emission from massless light-quark lines, introducing a soft singularity as and collinear singularities when or . Standard MS collinear subtraction at factorization scale absorbs the collinear divergences into the parton-to-photon fragmentation functions [31, 32]. The resulting physical photon spectrum at nonvanishing energy fraction is
where is the collinear-subtracted partonic photon spectrum, and are inclusive parton spectra () at energy fraction .
Counting , the NNLO spectrum consistently incorporates partonic hard coefficients through [32]. The fragmentation functions at are evaluated through four-flavor NLO time-like evolution starting from the Bourhis-Fontannaz-Guillet (BFG) set II parameterization [33].
To compare with experimental measurements while maintaining theoretical control, we restrict the analysis to the fiducial window , adopted by CLEO in fitting the Garcia i Tormo–Soto (GS) model [9] [1]. This kinematic choice excludes the nonperturbative soft-photon region () as well as the upper-endpoint regime (), where nonperturbative shape functions and threshold resummations become necessary. Within this window, the partial decay width is infrared and collinear safe:
In Figure 1(a), we compare our predicted differential spectra at LO, NLO, and NNLO with reconstructed CLEO data points derived from Fig. 15 of Ref. [1] using measured photon efficiencies, acceptance, and the total width [8]. The data points retain detector resolution effects. The NNLO correction markedly lowers the central prediction relative to NLO, bringing the theoretical spectrum into close agreement with CLEO observations across the bulk of the window . Near the upper boundary (), fixed-order perturbation theory produces negative bin averages for certain scale choices [Fig. 1](b), reflecting the growth of large endpoint logarithms that naturally delimits the domain of validity for fixed-order predictions.
Figure 1. Photon energy spectrum in radiative decay. (a) Bin-averaged differential width at the central scale compared with reconstructed CLEO data [1]. The NNLO error band indicates Monte Carlo integration uncertainties. (b) Renormalization-scale dependence over at fixed and fragmentation functions.
For the partial width within , the short-distance coefficients in Eq. (2) at the central scales are
Combining the analytic Born integral with the bin-averaged corrections yields the partial widths at the reference coupling:
The NNLO correction reduces the NLO rate by 41.6%. The Monte Carlo uncertainties account for all bin and channel covariances. Because the two-loop virtual spectrum is evaluated at discrete photon energy points via GK quadrature, an interpolation is required to integrate its contribution across the continuous fiducial window . Varying the choice of interpolation scheme induces a modest shift in of at most 0.00026 keV (0.06%), demonstrating that the associated systematic uncertainty is negligible. Under scale variation over [Fig. 2(a)], the NNLO radiative width spans 0.1290 to 0.5641 keV, corresponding to a relative scale uncertainty of , compared to at NLO.
Hadronic decay.—With identical input parameters, the inclusive hadronic annihilation widths are computed as
where the short-distance coefficients at the central scales are
Here quad denotes the quadrature error of the two-loop contribution. Converting to the four-flavor decoupling scheme [34] reproduces the established NLO coefficient [6, 35].
According to the PDG [8], the branching fraction is and the total width is (54.02 1.25) keV. Assuming uncorrelated uncertainties, the inferred experimental hadronic width is
Historically, the positive NLO correction pushed the theoretical width to 68.38 keV, substantially above the measured total width and creating a persistent tension. As shown in Figure 2(b), the large negative NNLO correction () reduces the prediction by over 32%, bringing into close agreement with the PDG reference value within uncertainties. The residual renormalization-scale uncertainty at NNLO is , reflecting the substantial magnitude of the two-loop coefficient.
Figure 2. Renormalization-scale dependence of the decay widths at the reference coupling . (a) The radiative window width , with the NNLO band showing Monte Carlo numerical error. (b) The inclusive hadronic width , with the NNLO band combining Monte Carlo error and the two-loop quadrature error. The experimental reference bands denote and the inferred radiative window width, with uncertainties propagated assuming uncorrelated inputs and a fixed GS window fraction.
Window ratio and determination of
Having established the two decay widths, we consider their ratio
in which the nonperturbative color-singlet matrix element and kinematic mass factors cancel identically. Because the Born widths scale as and , respectively, the ratio provides a direct handle on the strong coupling, scaling as at leading order.
We expand the theoretical ratio in powers of , retaining terms consistently through relative NNLO:
where and the perturbative coefficients are given by
c_{R,W}^{(2)}=c_{\gamma,W}^{(2)}-c_h^{(2)}+c_h^{(1)}\left(c_h^{(1)}-c_{\gamma,W}^{(1)}\right).
At fixed fragmentation input, substituting Eq. (3) into Eq. (13) reveals that the quadratic scale logarithms and the NRQCD factorization scale cancel identically, leaving
c_{R,W}^{(2)}=\bar{c}_{R,W}^{(2)}-\frac{\beta_1}{16}L_R.
The remaining linear logarithms compensate the running of the Born ratio through .
At the reference coupling , the expanded ratio normalized to its Born value is
The NNLO term is substantially smaller than the NLO correction . Under renormalization-scale variation , the relative uncertainty of the expanded ratio is reduced to [Figure 3(a)], exhibiting markedly enhanced perturbative stability compared to either individual width.
Figure 3. Window ratio at the input coupling (a) and extracted using the window ratio inferred from CLEO’s GS fit (b,c). The expanded NNLO bands show numerical errors in (a) and combined experimental and numerical errors in (b,c). At , the minimum unexpanded ratio for positive widths in the fit range is about 0.01149, above the CLEO central value 0.010365. The unexpanded curves begin at the next scale in the scan, .
For comparison, one may evaluate the unexpanded quotient . While agreeing with Eq. (12) through NNLO, it generates additional higher-order terms from expanding the denominator. At the reference coupling, the two prescriptions give
To connect with experiment, we utilize the CLEO measurement [1]. Fitting the background-subtracted photon spectrum to the GS model in yielded an extrapolated total ratio , with uncertainties and [28] [1]. Multiplying by the fraction of the GS spectrum within gives the fiducial window ratio:
Equating the expanded ratio with at central scale and evolving the coupling to at three loops yields
The experimental uncertainty from CLEO dominates the error budget. The numerical uncertainty originates from the Monte Carlo integration covariances of both widths and the hadronic quadrature error. The scale uncertainty is determined from the envelope of couplings fitted over and evolved to each scale.
Systematic sensitivities to nonperturbative and modeling inputs remain small: varying the reconstruction and endpoint extrapolation of the GS spectrum shifts by to , while switching between BFG-I and BFG-II and varying yield shifts below .
The expanded NNLO result is in good agreement with the world average [3] within experimental and numerical uncertainties. The unexpanded fit yields (Table I), a downward shift of 0.00783 that exceeds the expanded scale uncertainty Figure 3(c), indicating that uncalculated higher-order terms remain the leading source of theoretical systematic variation.
| Prediction | \multicolumn{4}{c}{} | \multicolumn{4}{c}{} | ||||||
| Central | data | num | scale | Central | data | num | scale | |
| LO | 0.24339 | — | 0.14104 | — | ||||
| NLO | 0.17850 | 0.11717 | ||||||
| NNLO expanded | 0.18868 | 0.12135 | ||||||
| NNLO unexpanded | 0.16998 | — | 0.11352 | — |
Table I. Values of extracted from the window ratio inferred from CLEO’s GS fit. Central values use . Scale uncertainties correspond to the envelope of fits evaluated across and evolved to the reference scale. Dashes indicate vanishing numerical error at LO. The scale uncertainty for the unexpanded fit is omitted because the quotient admits no physical solution at .
Conclusions. In summary, we have calculated the complete NNLO QCD corrections to both the radiative and inclusive hadronic annihilation decays of the meson in color-singlet NRQCD at leading order in the heavy-quark velocity expansion. In the radiative channel, the NNLO correction significantly reshapes the differential photon spectrum, bringing the central prediction into close agreement with reconstructed CLEO data across the fiducial window . In the hadronic channel, the substantial negative NNLO correction lowers the predicted decay width from its inflated NLO value, bringing it into full consistency with the PDG experimental reference and resolving a longstanding tension in vector quarkonium phenomenology.
Taking the ratio of the two widths eliminates the nonperturbative color-singlet matrix element and produces a stable perturbative series with suppressed renormalization-scale dependence. Applying this ratio to the CLEO measurement provides a clean determination of the strong coupling constant, , compatible with the world average. The shift between the expanded and unexpanded prescriptions highlights the importance of higher-order perturbative effects. Future high-luminosity data from Belle II, combined with endpoint resummations and complementary analyses of leptonic decays, will further advance precision tests of QCD in heavy quarkonium systems.
Acknowledgments
Support for this work was provided by the National Natural Science Foundation of China (Grant No. 12135013). Generative AI tools were used in developing code, checking analytic and numerical results, and preparing the manuscript.
References
References
- [2]N. Brambilla et al., Heavy quarkonium: progress, puzzles, and opportunities, Eur. Phys. J. C 71, 1534 (2011), arXiv:1010.5827 [hep-ph].DOI
- [3]Particle Data Group, Review of particle physics: Quantum chromodynamics (2026).
- [4]G. T. Bodwin, E. Braaten, and G. P. Lepage, Rigorous QCD analysis of inclusive annihilation and production of heavy quarkonium, Phys. Rev. D 51, 1125 (1995), erratum: Phys. Rev. D 55, 5853 (1997), arXiv:hep-ph/9407339.arxiv.org/abs/hep-ph/9407339
- [5]N. Brambilla, A. Pineda, J. Soto, and A. Vairo, Effective field theories for heavy quarkonium, Rev. Mod. Phys. 77, 1423 (2005), arXiv:hep-ph/0410047.DOI
- [6]P. B. Mackenzie and G. P. Lepage, Quantum chromodynamic corrections to the gluonic width of the ϒ meson, Phys. Rev. Lett. 47, 1244 (1981).
- [7]M. Krämer, QCD corrections to radiative Upsilon decays, Phys. Rev. D 60, 111503 (1999), arXiv:hep-ph/9904416.DOI
- [8]Particle Data Group, Review of particle physics: ϒ(1S) listing (2026).
- [9]X. Garcia i Tormo and J. Soto, Semi-inclusive radiative decays of ϒ(1S), Phys. Rev. D 72, 054014 (2005), arXiv:hep-ph/0507107.DOI
- [10]F. Feng, Y. Jia, and W.-L. Sang, Next-to-next-to-leading-order QCD corrections to the hadronic width of pseudoscalar quarkonium, Phys. Rev. Lett. 119, 252001 (2017), arXiv:1707.05758 [hep-ph].DOI
- [11]C. Zeng, B. Gong, J.-X. Wang, R. Niu, X.-D. Huang, and C. Li, Complete next-to-next-to-leading-order QCD correction to J/ψ → 3γ decay (2026), arXiv:2603.26199 [hep-ph].arxiv.org/abs/2603.26199
- [12]J. Soto, αₛ from hadronic quarkonia decays, in High-precision αₛ measurements from LHC to FCC-ee, edited by D. d’Enterria and P. Z. Skands (2015) arXiv:1512.05194 [hep-ph].
- [13]T. Hahn, Generating Feynman diagrams and amplitudes with FeynArts 3, Comput. Phys. Commun. 140, 418 (2001), arXiv:hep-ph/0012260.DOI
- [14]B. Ruijl, T. Ueda, and J. A. M. Vermaseren, FORM version 4.2 (2017), arXiv:1707.06453 [hep-ph].arxiv.org/abs/1707.06453
- [15]P. Maierhöfer, J. Usovitsch, and P. Uwer, Kira—a Feynman integral reduction program, Comput. Phys. Commun. 230, 99 (2018), arXiv:1705.05610 [hep-ph].
- [16]J. Klappert, F. Lange, P. Maierhöfer, and J. Usovitsch, Integral reduction with Kira 2.0 and finite field methods, Comput. Phys. Commun. 266, 108024 (2021), arXiv:2008.06494 [hep-ph].DOI
- [17]F. Lange, J. Usovitsch, and Z. Wu, Kira 3: Integral reduction with efficient seeding and optimized equation selection, Comput. Phys. Commun. 322, 109999 (2026), arXiv:2505.20197 [hep-ph].arxiv.org/abs/2505.20197
- [18]R. Niu, C. Zeng, B. Gong, and J. Wang, SecRed: A program for sector-wise IBP reduction of Feynman integrals (2026), arXiv:2611.03703 [hep-ph].arxiv.org/abs/2611.03703
- [19]X. Liu and Y.-Q. Ma, AMFlow: a Mathematica package for Feynman integrals computation via auxiliary mass flow, Comput. Phys. Commun. 283, 108565 (2023), arXiv:2201.11669 [hep-ph].DOI
- [20]L. Magnea, E. Maina, G. Pelliccioli, C. Signorile-Signorile, P. Torrielli, and S. Uccirati, Local analytic sector subtraction at NNLO, JHEP 12 (2018), 107, arXiv:1806.09570 [hep-ph].DOI
- [21]L. Magnea, G. Pelliccioli, C. Signorile-Signorile, P. Torrielli, and S. Uccirati, Analytic integration of soft and collinear radiation in factorised QCD cross sections at NNLO, JHEP 02 (2021), 037, arXiv:2010.14493 [hep-ph].DOI
- [22]G. Bertolotti, L. Magnea, G. Pelliccioli, A. Ratti, C. Signorile-Signorile, P. Torrielli, and S. Uccirati, NNLO subtraction for any massless final state: a complete analytic expression, JHEP 07 (2023), 140, arXiv:2212.11190 [hep-ph].DOI
- [23]F. Caola, K. Melnikov, and R. Röntsch, Nested soft-collinear subtractions in NNLO QCD computations, Eur. Phys. J. C 77, 248 (2017), arXiv:1702.01352 [hep-ph].DOI
- [24]F. Caola, K. Melnikov, and R. Röntsch, Analytic results for decays of color singlets to gg and qq̄ final states at NNLO QCD with the nested soft-collinear subtraction scheme, Eur. Phys. J. C 79, 1013 (2019), arXiv:1907.05398 [hep-ph].arxiv.org/abs/1907.05398
- [25]S. Catani, The singular behaviour of QCD amplitudes at two-loop order, Phys. Lett. B 427, 161 (1998), arXiv:hep-ph/9802439.DOI
- [26]Z. Bern, L. J. Dixon, and D. A. Kosower, Two-loop g → gg splitting amplitudes in QCD, JHEP 08 (2004), 012, arXiv:hep-ph/0404293.DOI
- [27]E. J. Eichten and C. Quigg, Quarkonium wave functions at the origin, Phys. Rev. D 52, 1726 (1995), arXiv:hep-ph/9503356.DOI
- [28]N. Brambilla, X. Garcia i Tormo, J. Soto, and A. Vairo, Extraction of αₛ from radiative ϒ(1S) decays, Phys. Rev. D 75, 074014 (2007), arXiv:hep-ph/0702079.
- [29]M. Beneke, A. Signer, and V. A. Smirnov, Two-loop correction to the leptonic decay of quarkonium, Phys. Rev. Lett. 80, 2535 (1998), arXiv:hep-ph/9712302.DOI
- [30]H. S. Chung, MS renormalization of S-wave quarkonium wavefunctions at the origin, JHEP 12 (2020), 065, arXiv:2007.01737 [hep-ph].DOI
- [31]S. Catani and F. Hautmann, Surprises from quarkonium decay into photons, Nucl. Phys. B Proc. Suppl. 39BC, 359 (1995), arXiv:hep-ph/9410394.DOI
- [32]T. Gehrmann and R. Schürmann, Photon fragmentation in the antenna subtraction formalism, JHEP 04 (2022), 031, arXiv:2201.06982 [hep-ph].DOI
- [33]L. Bourhis, M. Fontannaz, and J. P. Guillet, Quark and gluon fragmentation functions into photons, Eur. Phys. J. C 2, 529 (1998), arXiv:hep-ph/9704447.DOI
- [34]K. G. Chetyrkin, B. A. Kniehl, and M. Steinhauser, Decoupling relation to O(αₛ³) and their connection to low-energy theorems, Nucl. Phys. B 510, 61 (1998), arXiv:hep-ph/9708255.DOI
- [35]W.-L. Sang, F. Feng, and Y. Jia, O(αₛv²) corrections to hadronic decay of vector quarkonia, Phys. Rev. D 102, 094021 (2020), arXiv:2010.11896 [hep-ph].DOI


