A collaborative effort between Harvard University and scientists at QuEra Computing, University of Innsbruck, MIT, and other institutions has led to the demonstration of a breakthrough application of neutral-atom quantum processors for solving problems of practical use.
Earlier, neutral-atom quantum processors have been proposed for efficient encoding of certain difficult combinatorial optimization problems. The landmark publication from the authors explains first time implementation of efficient quantum optimization on a real quantum computer, and also showcases unprecedented quantum hardware power.
Harvard’s quantum processor of 289 qubits operating in analog mode were used for calculations with effective circuit depths up to 32. Importantly, the circuit depth and large system size used in this work made it impossible to employ classical simulations for pre-optimizing of control parameters unlike previous works of quantum optimization. This involved deploying a quantum-classical hybrid algorithm in a closed loop which directly provided automated feedback to the quantum processor.
The combination of circuit depth, system size, and outstanding quantum control resulted in a quantum leap: Problem areas were found with performance of quantum processor which is better than expected empirically versus classical heuristics. The characterization of difficulty of optimization problem occurrence with a hardness parameter led the team to identify cases that challenged classical computers, but were more efficiently solved with neural-atom quantum processors.
This led to finding a super linear quantum speed compared to a class of generic classical algorithms. The open source algorithms GenericTensorNetworks.jl and Bloqade.jl of QuEra were instrumental to discover hard instances and understanding quantum performance.
A detailed understanding of underlying physics of quantum algorithms and fundamental constraints of its classical equivalent allowed to realize methods for quantum machine to attain a speedup.
