DeepQuark: A Deep-Neural-Network Approach to Multiquark Bound States

Area of Science:

• Quantum Chromodynamics (QCD)
• Nuclear Physics
• Computational Physics

Background:

• Multiquark systems present significant computational challenges due to strong SU(3) color interactions and complex correlations.
• Existing methods struggle with the intricacies of confinement and extra quantum numbers in these systems.

Purpose of the Study:

• To develop and implement a novel deep neural network-based variational Monte Carlo approach for multiquark bound states.
• To address the computational barriers and unique challenges posed by multiquark systems.
• To provide a powerful framework for exploring confining mechanisms beyond two-body interactions.

Main Methods:

• Implementation of a deep-neural-network-based variational Monte Carlo (VMC) approach.
• Design of a novel, high-efficiency architecture named DeepQuark.
• Incorporation of three-body flux-tube confinement interactions without additional computational cost.

Main Results:

• DeepQuark demonstrates competitive performance against state-of-the-art methods for nucleon and tetraquark systems.
• The approach achieves superior accuracy for pentaquark systems, including triply heavy pentaquarks.
• Prediction of weakly bound D*Ξcc* molecule Pccc¯(5715) and its bottom analog Pbbb¯(15569).

Conclusions:

• DeepQuark offers a promising solution for studying larger multiquark systems, overcoming computational limitations.
• The framework provides insights into nonperturbative QCD and many-body physics by exploring confinement mechanisms.
• Experimental searches for predicted pentaquark states, such as Pccc¯(5715), are recommended.