Quantum machine-learning phase prediction of high-entropy alloys

Discovering new high-entropy alloys (HEAs) in the vast compositional space requires a growing power of classical computers for training machine learning models. The exponential increase of HEA data will pose a challenge in making the machine learning process prohibitively time consuming in the foreseeable future. Quantum computers, which use quantum superposition and interference to perform computations, hold great potential in handling big data and accelerating the optimization algorithms ubiquitous in machine learning models. Here we adopt a quantum computer simulator and quantum processors to prepare for the future challenge in new HEA discovery. We first train a classical artificial neural network (ANN), which uses HEAs' compositions as inputs and the corresponding phases as outputs, to predict phase selection. We then apply a quantum computer simulator that implements a hybrid quantum–classical machine learning algorithm to accomplish the same supervised machine learning task. We find that the resulting testing accuracy is comparable to that from classical ANN calculations. We finally apply quantum processors to perform the hybrid quantum–classical machine learning calculations and obtain slightly lower accuracy ascribed to the fragile nature of quantum bits in quantum processors. Our work initiates the adoption of fledgling quantum computers in the noisy intermediate-scale quantum (NISQ) era for discovering new HEAs.