Abstract
We introduce a genetic algorithm that designs quantum optics experiments forengineering quantum states with specific properties. Our algorithm is powerfuland flexible, and can easily be modified to find methods of engineering statesfor a range of applications. Here we focus on quantum metrology. First, weconsider the noise-free case, and use the algorithm to find quantum states witha large quantum Fisher information (QFI). We find methods, which only involveexperimental elements that are available with current or near-futuretechnology, for engineering quantum states with up to a 100-fold improvementover the best classical state, and a 20-fold improvement over the optimalGaussian state. Such states are a superposition of the vacuum with a largenumber of photons (around $80$), and can hence be seen asSchr\"odinger-cat-like states. We then apply the two most dominant noisesources in our setting -- photon loss and imperfect heralding -- and use thealgorithm to find quantum states that still improve over the optimal Gaussianstate with realistic levels of noise. This will open up experimental andtechnological work in using exotic non-Gaussian states for quantum-enhancedphase measurements. Finally, we use the Bayesian mean square error to lookbeyond the regime of validity of the QFI, finding quantum states with precisionenhancements over the alternatives even when the experiment operates in theregime of limited data.