Trieste Quantum Seminars:

Physics-Informed Learning and Digital Quantum Simulation of Quantum Critical Dynamics

Dr. Antonio Mandarino
International Centre for Theory of Quantum Technologies, University of Gdansk, Poland

Abstract: Machine learning and quantum simulation are becoming complementary pillars for exploring quantum many-body physics beyond the reach of exact analytics and classical computation. On the one hand, modern learning methods can extract structure from the quantum formulation of the problem under scrutiny even when labels are scarce or unavailable. On the other hand, increasingly accessible quantum processing units (QPUs) provide a programmable experimental platform to generate and validate quantum-dynamical phenomena directly, turning quantum hardware into a broadly available laboratory. In this seminar, I illustrate how these two directions meet in the study of quantum criticality. To address the bottleneck of labeled data in supervised phase recognition, we employ Physics-Informed Neural Networks (PINNs) in an unsupervised setting, using physical constraints to guide learning in models ranging from the anharmonic oscillator to critical spin chains. In parallel, we use gate-based quantum hardware as a digital quantum simulator to probe universal scaling in driven critical systems. Focusing on quench dynamics, we measure the defect density as a function of quench time and test the Kibble–Zurek mechanism in finite size devices, highlighting how hardware noise and limited system size reshape, but do not erase universal trends. This hybrid workflow points toward scalable strategies for mapping phase structure and nonequilibrium universality in increasingly complex quantum many-body systems.