Physics-Informed deep operator learning framework for multiphysics heat and mass transport in rotating disk energy systems

• Introduces a physics-informed deep operator learning framework. • Models coupled heat–mass transport in rotating MHD Reiner–Rivlin flows. • Achieves <0.5 % mean error vs. high-fidelity finite element solutions. • Provides 100 × faster prediction of Nusselt and Sherwood metrics. • Enables real-time design and control of rotating thermal energy systems. Accurate prediction of coupled heat and mass transport in rotating energy systems remains a challenge due to the nonlinear interaction between magnetic, viscous, and diffusive mechanisms. This study introduces a physics-informed deep operator learning framework for modeling multiphysics flow and transport in Reiner–Rivlin fluids subjected to magnetohydrodynamic (MHD) effects, viscous dissipation, and cross-diffusion through Soret and Dufour phenomena. The hybrid methodology combines high-fidelity finite element reference solutions with deep operator networks (DeepONet) trained under physics-guided loss constraints, enabling generalization across wide parametric domains. Results show that the proposed model captures the interplay between Lorentz damping, viscous heating, and thermo-diffusion with high quantitative fidelity, achieving mean absolute errors of <0.5 % for velocity, temperature, and concentration profiles compared to numerical baselines. The surrogate efficiently predicts wall transfer metrics such as the Nusselt and Sherwood numbers across broad variations in magnetic intensity, rheological parameters, and volumetric heating rates, with a computational speed-up exceeding two orders of magnitude relative to conventional solvers. The framework demonstrates how physics-aware learning can bridge numerical accuracy and real-time predictability, providing a robust foundation for the design and optimization of rotating thermal systems.