Quantum Dark Magic: Efficiency of Intermediate Non-Stabiliserness

Abstract: While superiority of quantum over classical computation has been established, the repertoire of primitives with proven or conjectured quantum advantage remains limited. Despite considerable progress in delineating the quantumclassical divide, the systematic construction of algorithms with quantum advantage remains challenging, which can be attributed to a still incomplete understanding of the sources of quantum computational power. While intermediate non-stabiliserness (i.e., traversal of states outside the Clifford orbit) indicates necessary non-classical behaviour for quantum advantage, naively equating non-stabiliserness and non-classicality is misguided: Even random Haar sampled states exhibit near-maximal non-stabiliserness. Advancing towards quantum advantage calls for a better understanding of the efficient use of non-stabiliser states. We present an approach to track the behaviour of non-stabiliserness across various algorithms by pairing resource theory of non-stabiliser entropies with the geometry of quantum state evolution, and introduce permutation agnostic distance measures that reveal non-stabiliser effects previously hidden by a subset of Clifford operations. We find different efficiency in the use of non-stabiliserness for structured and unstructured variational approaches, and show that greater freedom for classical optimisation in quantum-classical methods increases unnecessary non-stabiliser consumption. Our results open new means of analysing the efficient utilisation of quantum resources, and contribute towards the targeted construction of algorithmic quantum advantage.