Digital Annealing Route to Complex Magnetic Phase Discovery

Emerging computational paradigms and device architectures may provide more robust, efficient routes for scientific discovery compared to traditional architectures. One such paradigm is that of the Digital Annealer (DA), a hardware-accelerated device designed to implement Markov Chain Monte Carlo algorithms with lower overhead than traditional device architectures. To better understand the applicability of digital annealing for scientific discovery, we explore the application of the fully connected 8000 variable Fujitsu DA for a complex material science problem. We identify the intricate phases and the phase transitions of an Ising model defined on an extended Shastry-Sutherland lattice, which is believed to effectively describe the magnetic physics in a host of potential spintronic materials. To validate our implementation, we identify all previously known solutions to the model, including the nontrivial and highly non-degenerate 1/3,1/2,1/5, and 5/9 fractional magnetization plateaus. Accounting for the boundary effects, we find that the Fujitsu DA provides immaculate quality of solutions, even close to a phase transition where classical Monte Carlo codes can often struggle to converge. We then take advantage of the full connectivity of the DA, and its tunable parameters to discover new phases and their interesting spin motifs not previously known. We conclude that digital annealing provides a novel route for discovery of complex magnetic phases, opening avenues for the understanding and engineering of spintronics materials.