Efficient Synthesis of Probabilistic Quantum Circuits with Fallback

Recently it has been shown that Repeat-Until-Success (RUS) circuits can approximate a given single-qubit unitary with an expected number of T gates of about 1/3 of what is required by optimal, deterministic, ancilla-free decompositions over the Clifford+T gate set. In this work, we introduce a more general and conceptually simpler circuit decomposition method that allows for synthesis into protocols that probabilistically implement quantum circuits over several universal gate sets including, but not restricted to, the Clifford+T gate set. The protocol, which we call Probabilistic Quantum Circuits with Fallback (PQF), implements a walk on a discrete Markov chain in which the target unitary is an absorbing state and in which transitions are induced by multi-qubit unitaries followed by measurements. In contrast to RUS protocols, the presented PQF protocols terminate after a finite number of steps. Specifically, we apply our method to the Clifford+T, Clifford+V , and Clifford+π/12 gate sets to achieve decompositions with expected gate counts of logb(1/ε)+O(log(log(1/ε))), where b is a quantity related to the expansion property of the underlying universal gate set.