Dynamic thermalization on noisy quantum hardware

Digital quantum simulation is a promising tool for studying thermalization in quantum many-body systems, providing unprecedented control but remaining limited by sparse resources of noisy intermediate-scale quantum devices (NISQ). We demonstrate a dynamical thermalization mechanism based on averaging the quantum mechanical observables over randomized variants of their time evolution,which does not rely on a system's size or the presence of a thermal bath. As the evolution begins, average occupation probabilities of many-body energy states quickly equilibrate toward the Gibbs distribution with a finite temperature depending on the initial state's energy. Running an experiment on an IBM Quantum computer for four qubits, we report the utility of a NISQ device for predicting thermal observables and their fluctuations and demonstrate the dynamical emergence of an equilibrium state with a negative absolute temperature.