We present an exact synthesis algorithm for qutrit unitaries in
U3n(Z[1/3,e2πi/3]) over the Clifford
+T gate set with at most one ancilla. This extends the already known result of qutrit metaplectic gates being a subset of Clifford
+T gate set with one ancilla. As an intermediary step, we construct an algorithm to convert 3-level unitaries into multiply-controlled gates, analogous to Gray codes converting 2-level unitaries into multiply-controlled gates. Finally, using catalytic embeddings, we present an algorithm to exactly synthesize unitaries
U3n(Z[1/3,e2πi/9]) over the Clifford
+T gate set with at most 2 ancillas. This, in particular, gives an exact synthesis algorithm of single-qutrit Clifford
+D over the multi-qutrit Clifford
+T gate set with at most two ancillas.
Phase gadgets have proved to be an indispensable tool for reasoning about ZX-diagrams, being used in optimisation and simulation of quantum circuits and the theory of measurement-based quantum computation.In this paper we study phase gadgets for qutrits.We present the flexsymmetric variant of the original qutrit ZX-calculus, which allows for rewriting that is closer in spirit to the original (qubit) ZX-calculus.In this calculus phase gadgets look as you would expect, but there are non-trivial differences in their properties.We devise new qutrit-specific tricks to extend the graphical Fourier theory of qubits, resulting in a translation between the 'additive' phase gadgets and a 'multiplicative' counterpart we dub phase multipliers.This enables us to generalise the qubit notion of multiple-control to qutrits in two ways.The first type is controlling on a single tritstring, while the second type applies the gate a number of times equal to the tritwise multiplication modulo 3 of the control qutrits.We show how both types of control can be implemented for any qutrit Z or X phase gate, ancilla-free, and using only Clifford and phase gates.The first requires a polynomial number of gates and exponentially small phases, while the second requires an exponential number of gates, but constant sized phases.This is interesting, because such a construction is not possible in the qubit setting.As an application of these results we find a construction for emulating arbitrary qubit diagonal unitaries, and specifically find an ancilla-free emulation for the qubit CCZ gate that only requires three single-qutrit non-Clifford gates -provably lower than the four T gates needed for qubits with ancilla.