Research Projects

The central challenge to be taken up in this project is to describe the finite temperature properties of quantum many-body systems using artificial intelligence(AI)-neural network (NN)-based methods. The present investigators propose a new methodology based on NN-based neural networks that takes inspiration from genetic algorithms to study quantum many-body systems at finite temperature. The project will be organized into three subheads. 1) Can NN-based states provide an efficient framework to study quantum systems at the experimentally relevant regime of finite temperatures? The project proposes a new methodology for the simulation of quantum many-body systems – the method uses a family of states which are progressively sampled to obtain a description of the statistical thermal ensemble of the system. The method exploits the power of NN-states to represent states with a larger amount of entanglement, and the power of GPU computations to manipulate the family of states. The method will be developed to study theoretical models that simulate quantum memories, such as Z2 gauge theories on random graphs (to localize the monopole excitations). 2) How to understand the neural-network states from the perspective of the renormalization group? Can some features of the neural states be identified with relevant and irrelevant operators near a critical point (either classical or quantum)? The project aims to bring to bear renormalization group ideas to understand the information encoded in neural network states that represent various quantum states. This will be explored based on a coarse-graining strategy of neural states that will find the key parameters in the neural states that characterize the phases describing the neural states. 3) Can these neural states teach us something that is not possible with established methods? We propose to explore Z2 gauge theories with disorder, which are candidates for stable quantum memories and are very difficult to tackle with established methods, using our NN-based method.

Neural-Network Many-Body State, Finite Temperature Algorithms, Topologically Ordered States, AI methods