QUACE: Quantum Atomic Cluster Expansion

Speaker: Lila Cadi Tazi
University of Cambridge
QUACE: Symmetrized molecular descriptors on a quantum circuit
There is an evident need to develop fast and accurate methods for materials and chemical simulations, as ab initio quantum chemistry methods are not scalable. On the one hand, the machine learning (ML) literature offers an appealing approach in which ML force fields trained on a small number of ab initio computations can accurately predict properties of new structures with a reduced computational cost. On the other hand, quantum computing algorithms are being developed with the potential to reduce the scaling of classical computing methods for theoretical chemistry.
In this work, we bring together quantum computing and the MACE ML framework to introduce the QUACE hybrid algorithm.
In the classical MACE architecture, the central operation is a tensor contraction that requires the manipulation of highly dimensional data and is the bottleneck step of the method. This tensor operation is well suited to be performed on a quantum circuit.
We implement a quantum algorithm that performs tensor contraction, thus reducing the load of classical processing and enabling improved scaling of the overall algorithm with system size.
Although present NISQ quantum devices are not powerful enough to outperform classical performances, this work aims to demonstrate the potential of quantum computing for tensor contractions and its application to molecular simulations.
QUACE could be a practical method for running molecular dynamics on near-term noisy quantum devices, where quantum noise would be harnessed as a source of stochasticity in the dynamics.