Seminarium Fizyki Ciała Stałego

The growing limitations of conventional electronics are driving the search for alternative computing platforms capable of ultra-fast and energy-efficient information processing. Exciton-polaritons are promising candidates for optical neuromorphic devices owing to their hybrid light-matter nature, which combines strong optical nonlinearity with efficient transport. However, most polariton neural network implementations so far have relied on cryogenic operation, limiting their practical deployment. Here, we present a major step toward room-temperature polaritonic computing based on large-scale CsPbBr₃ perovskite microwires with arbitrary in-plane geometries. Using a versatile template-assisted fabrication method, we create high-optical-quality waveguiding microstructures that do not require conventional cavity mirrors, greatly simplifying device fabrication and improving compatibility with integrated photonic architectures. These microwires support room-temperature exciton-polariton condensation and exhibit polariton lasing with pronounced blueshifts at high excitation densities, whereas emission from the edges of the structure provides evidence of long-range polariton transport. At the same time, TE–TM splitting of the waveguided modes induces optical spin–orbit coupling, which enables polarization control of propagating polariton packets and thereby supports spin-polarized transport at room temperature. Importantly, these perovskite waveguides are exceptionally well suited for integration into integrated photonic platforms. By implementing subwavelength grating couplers, we demonstrate efficient and controllable in- and out-coupling of light to guided exciton-polariton modes without compromising the crystalline quality of the microwires. These combined properties position perovskite microwires as promising building blocks for integrated photonic architectures, reconfigurable polaritonic circuits, and photonic neural networks, in which they can serve as individual or coupled nonlinear nodes. We further discuss the potential of this platform for machine learning tasks such as binary classification and object recognition. With its simple fabrication, scalability, room-temperature operation, and compatibility with integrated photonic platforms, this system offers a promising route toward practical polaritonic computing devices.