How Often Must Quantum Error Correction be Implemented?

Correcting errors is a vital but expensive component of fault tolerant quantum computation. Standard fault tolerant protocol assumes the implementation of error correction, via syndrome measurements and possible recovery operations, after every quantum gate. In fact, this is not necessary. Here we demonstrate that error correction should be applied more sparingly. We simulate encoded single qubit rotations within the [7,1,3] code and show via fidelity measures that applying error correction after every gate is not desirable. The simulations also shed light on what accuracy can be expected for noisy error correction and thus to what accuracy arbitrary single qubit rotations should be implemented.