Q-ONE aims at exploring a novel approach for sensing and generating quantum states of light. The project idea places itself at the frontier between quantum physics and applied artificial intelligence. The consortium targets the realization of a novel device based on strongly interacting photons (exciton-polaritons) that, using principles of neuromorphic computing, is able to recognize, characterize, and generate a variety of quantum states.
Q-ONE aims at exploring a novel approach for sensing and generating quantum states of light. The project idea places itself at the frontier between quantum physics and applied artificial intelligence. The consortium targets the realization of a novel device based on strongly interacting photons (exciton-polaritons) that, using principles of neuromorphic computing, is able to recognize, characterize, and generate a variety of quantum states. Photons are the best particles to use for quantum application giving their robustness to decoherence and for their relatively easy generation of quantum states. However, one of the major drawbacks is the very small nonlinearities that photons can feel in standard nonlinear media. In this proposal we will make use of a hybrid state of light and matter, the exciton-polariton.
Exciton-polaritons are very interesting quantum quasiparticles that can find potential applications in various fields, such as extremely accurate interferometric measurements, ultra-low power lasers or information processing with very low energy losses. Exciton-polaritons are formed in semiconductor materials with a specially designed structure, thanks to the strong coupling of photons and excitons, which are material particles composed of electrons and “holes”. Polaritons possess a “Schrödinger cat” structure. The quantum state contains two alternatives: cat alive when the exciton exists, or dead cat when instead of an exciton a photon exists in the system. In 2006 came the news that condensation of exciton-polaritons was achieved in a specially prepared semiconductor sample. The idea of Bose-Einstein condensation was sparked in the mid-twenties of the last century, when Albert Einstein, inspired by the Indian physicist Satyendra Nath Bose working on the statistical properties of optical radiation, predicted a particular type of phase transition in the system of non-interacting bosons at temperature close to the absolute zero. In 1938, Fritz London linked this phenomenon with the recently discovered superfluidity of helium, thereby pointing to the first practical application of this rather abstract concept. The observation of a Bose-Einstein condensate of rubidium and sodium gases was achieved much later, in 1995, opening a new chapter in the field of atomic and molecular physics and quantum optics. But only when the condensation of exciton-polaritons was achieved, it could be realized at room temperature, bringing it much closer to real-life applications. Imagine a crowd of people, each person going in its own direction, sometimes bouncing against the others. This can be thought of as a model of a classical as of particles in a typical thermal state. Now consider an army of soldiers marching at the same pace, moving in unison. This will be the picture of the Bose-Einstein condensate, a quantum gas of bosons at low temperature. From time to time, however, one of the soldiers will do something unexpected, which will break the complete coherence of their movement. This is what happens when quantum fluctuations come into play, which are in fact unavoidable in quantum theory once we consider the individual movements of particles.
Biological neural networks achieve unmatched power in performing certain tasks, despite the relatively slow response time (millisecond scale) of their individual elements. Consequently, they have inspired research into artificial neural networks (NNs), which are particularly efficient at finding correlations and recognizing patterns in complex data. Here, we propose to use NNs to solve problems of quantum physics, starting with the recognition of patterns in quantum data corresponding to quantum state tomography. Unlike existing schemes, our NN approach would avoid the time-consuming procedure of multiple measurements typically used in quantum state tomography by basing all its interpretation on occupation numbers that inherit signatures from the input quantum states. This allows the recognition of a quantum state simply from the measurement of the output intensities of the quantum reservoir processor.
In this project, though, we don’t need to enter the full quantum regime using extraordinary interactions since we propose to exploit the properties of a quantum neural networks (QNN) of polariton nonlinear nodes, using state-of-the-art interactions, to identify and generate quantum states: this strongly innovative idea relies on the resonant injection of states as excitations of the QNN, which physically realizes–rather than simulates–a massively parallel computing task. Indeed, based on recent theoretical advancements, prompting the feasibility of quantum neural network with presently achieved polariton interactions and modal volumes, we propose to implement different polariton platforms to solve one of the most interesting problems of quantum mechanics: the recognition of quantum states of photons, like Fock states or entangled pairs, without the need of correlation measurements (like those in quantum tomography). Moreover, we will explore the idea of converting classical light from a classical source into a quantum state.