Heavy Quarkonium Spectrum and Decay Constants from a Neural-Network-Based Holographic Model

Yu Zhang, Xun Chen, Miguel Angel Martin Contreras · 2026

Read Heavy Quarkonium Spectrum and Decay Constants from a Neural-Network-Based Holographic Model by Yu Zhang, Xun Chen, Miguel Angel Martin Contreras (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-18133.txt.

Introduction

Yu Zhang, Xun Chen, and Miguel Angel Martin Contreras let a neural network draw the field that shapes heavy quark states. Their 2026 preprint builds a holographic model of heavy vector quarkonia. In the bottom-up AdS/QCD approach, a dilaton field sets the rules of the model. Most work picks a simple formula for that field by hand. The authors instead train a multilayer perceptron to learn the field's slope as a smooth function of the holographic coordinate.

A learned field can fit data without the bias of a guessed form. From it, the team computes the spectrum and decay constants of the states. The method joins machine learning with a theory of strong forces. It may sharpen predictions for particles that bind a heavy quark pair.

Catalog record

CREATORYu Zhang, Xun Chen, Miguel Angel Martin Contreras
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE32 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-18133

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Heavy Quarkonium Spectrum and Decay Constants from a Neural-Network-Based Holographic Model Yu Zhang, Xun Chen, Miguel Angel Martin Contreras Abstract We present a data-driven inverse construction of the dilaton field in a bottom-up AdS/QCD description of heavy vector quarkonia. Instead of adopting an ad hoc analytic ansatz, we use a multilayer perceptron to learn Φ ′ ​ ( z ) \Phi^{\prime}(z) as a smooth function of the holographic coordinate, with Φ ⁡ ( 0 ) = 0 \Phi(0)=0 imposed to ensure ultraviolet consistency. The dilaton and its derivatives obtained by automatic differentiation generate the holographic potential U ⁡ ( z ) U(z) , and the associated Schrödinger-like equation is discretized and diagonalized to extract the low-lying eigenmodes. Masses and decay constants are then evaluated from the eigenvalues and the near-boundary behavior of the bulk-to-boundary modes. Training on PDG data for charmonium and bottomonium yields a non-quadratic dilaton profile that resolves the longstanding difficulty of simultaneously reproducing both the heavy-quarkonium spectrum and the monotonic suppression of leptonic decay constants with radial excitation. The combined fit achieves RMS deviations of 1.26 % 1.26\% (charmonium) and 3.32 % 3.32\% (bottomonium). This work establishes neural-network reconstruction as a flexible tool for holographic modeling and provides a basis for future extensions incorporating additional channels, lattice constraints, or finite-temperature backgrounds. I Introduction Quantum Chromodynamics (QCD) is the fundamental theory that describes the strong interaction. However, in the low-energy and strongly coupled regime, analytical solutions to QCD become extremely difficult to obtain, and traditional perturbative methods are no longer effective. This limitation has motivated the development of various non-perturbative approaches, among which the AdS/QCD model 70 ; 13 ; 32 ; 18 based on the holographic principle has attracted considerable attention in recent years. The idea of AdS/QCD originates from the AdS/CFT correspondence, which reveals a profound connection between strongly coupled gauge theories and weakly coupled gravity theories 58 ; 5 ; 37 . Depending on the construction scheme, AdS/QCD models can be classified into top-down and bottom-up approaches: the former are derived from string theory, where probe branes are introduced into the bulk geometry to set the energy scale and break conformal symmetry 47 ; 74 ; 75 ; the latter directly implement deformations in the five-dimensional spacetime to reproduce hadronic properties observed in experiments. Within the bottom-up framework, the hard-wall 70 ; 13 ; 14 and soft-wall 46 models are the most widely used. In particular, the soft-wall model is well-suited for studying meson radial excitations, as it naturally reproduces the linear Regge trajectories of hadron masses. These models have been extensively applied to the study of meson spectroscopy 46 ; 35 ; 31 ; 26 ; 10 ; 28 ; 51 ; 85 ; 20 ; 81 and heavy quarkonia 49 ; 34 ; 53 ; 17 . The latter have introduced quadratic-like dilaton fields to fit heavy vector quarkonia with linear radial Regge trajectories. However, for heavy quarkonium systems, the radial Regge trajectory hypothesis breaks down, according to Bethe-Salpeter Theory 23 ; 24 ; 25 , implying Regge trajectories can be parametrized as M_{n}^{2}=a(n+b)^{\nu} . This non-linearity, at the holographic level, can be translated into the dilaton field \Phi(z) if a non-quadratic version is considered, as it was introduced in 62 . Recently, this dilaton phenomenological ansatz was derived from WKB analysis 64 . On the purely phenomenological side, heavy quarkonia, including charmonium ( c\,\bar{c} ) and bottomonium ( b\,\bar{b} ), play a central role in hadron physics due to their large quark masses, which simplify their internal dynamics. These systems provide an effective framework for investigating the non-perturbative regime of QCD. Heavy quarkonia connect perturbative calculations with strong-coupling phenomena, enabling rigorous tests of fundamental QCD concepts such as confinement and mass generation. Systematic analyses of their mass spectra and decay constants yield valuable information about the quark-antiquark potential. Furthermore, the well-characterized properties of heavy quarkonia serve as essential benchmarks for theoretical models, including lattice QCD and holographic methods, that seek to describe the strong interaction. Experimentally, the decay constants of charmonium and bottomonium can be extracted from the heavy vector to dileptons ( V\to l\,\bar{l} ) decay width. These constants display a monotonic decrease with increasing excitation level 65 . This trend indicates that higher excited states are progressively less stable 15 ; 63 , a defining feature of heavy quarkonium spectroscopy. In contrast, the decay constants of light mesons, such as the \rho , \phi , and \omega mesons, are only known experimentally for their ground states 65 . However, from theoretical approaches such as Relativistic Hamiltonian Dynamics 50 , Light-Front Dynamics 9 ; 82 , or the Nambu-Jona-Lasinio Model 21 , one can infer the decay constants of excited states. These theoretical estimates seem to show a decrease in decay constants with increasing radial excitation level, suggesting that this pattern may be universal in meson spectroscopy. It should be emphasized that, for heavy–light mesons such as D and B, the decay constants are extracted from electroweak processes rather than electromagnetic ones. Consequently, the underlying mechanism differs from that of heavy quarkonia, where decay constants are obtained from electromagnetic decays into dileptons. However, the existing holographic models still have limitations in describing heavy quarkonium spectroscopy, as they cannot simultaneously reproduce both the mass spectrum and the decay constants with high accuracy. For example, Ref. 34 improved the masses and decay constants of J/\psi and \psi^{\prime} by modifying the holographic potential, but failed to reproduce the experimental observation that the decay constants decrease with increasing excitation level. Refs. 33 ; 3 ; 4 ; 17 introduced a finite z_{uv} cutoff in the AdS slice to set the ultraviolet scale, which led to a decreasing trend of decay constants with higher excitation levels. However, the slope of this decrease was much smaller than that observed in the experimental results, providing only qualitative agreement. Another class of holographic models proposed in Ref. 16 successfully addressed the slope problem of the decay constants and accurately reproduced the spectra of charmonium and bottomonium decay constants. However, they could not simultaneously provide a precise description of the mass spectrum. In Refs. 61 ; 59 , the authors inferred the role played by the ultraviolet (UV) limit of the dilaton field \Phi(z) in determining decay constants. Therefore, modifying this limit allows us to induce the expected decrease in decay constants. Thus, transformations like isospectrality 60 or geometric deformations (as UV cutoffs) also modify decay constants. Recently, machine learning techniques, particularly multilayer perceptrons (MLPs), have shown significant promise in addressing complex scientific challenges 57 ; 90 ; 69 ; 54 ; 72 ; 68 ; 80 ; 84 ; 44 ; 41 ; 40 . The universal approximation theorem 43 establishes that an MLP can approximate any continuous function, provided a sufficient number of hidden neurons are present. This capability renders MLPs highly effective for solving partial differential equations 73 ; 78 ; 48 ; 76 and modeling inverse problems. Compared to traditional polynomial fitting or parametrization approaches, MLPs offer greater expressive power and enhance predictive accuracy while maintaining physical constraints. To the best of our knowledge, this work represents the first application of MLPs to reconstruct the dilaton field directly \Phi(z) within a bottom-up AdS/QCD model. Unlike previous

Read the complete text