Non-Clifford Gate on Gottesman-Kitaev-Preskill Encoded Optical Qubits with Nonlinear Feedforward

Achieving a fault-tolerant universal quantum computer with an optical system is one of the goals of quantum optics. For that purpose, the most promising candidate is the Gottesman-Kitaev-Preskill (GKP) encoding of qubits in the continuous-variable system [1] . In order to perform universal computation with GKP qubits, the current open problem is the realization of non-Clifford operations, which require non-Gaussianity. In the original GKP’s paper where [1] , two different approaches are suggested to implement a non-Clifford T gate, i.e. $\hat T = \left| {{0_{\text{L}}}} \right\rangle \left\langle {{0_{\text{L}}}\left| { + {{\text{e}}^{i\frac{\pi }{4}}}} \right|{1_{\text{L}}}} \right\rangle \left\langle {{1_{\text{L}}}} \right|$ , where |0 L 〉 and |1 L 〉 are the logical bases of the GKP qubits. The first approach is using a cubic phase gate, a non-Gaussian operation, combined with Gaussian operations. In this context, the cubic phase gate has been widely developed with nonlinear feedforward as a key technology [2] , [3] . In recent research, however, it is shown that this approach is not optimal for GKP qubits [4] . Hence, the second approach, which is based on the magic state injection method [5] , is more promising.