Silicon and Neuromorphic Photonics
Neuromorphic processors are widely considered one of the next frontiers in computing. We are fundamentally investigating information processing by leveraging the strengths of nanophotonic (i.e., optical) physics and neuromorphic (i.e., brain-inspired) architectures, and emerging technology (silicon and graphene photonics) platforms. We seek to develop programmable photonic processors that have the potential to outperform state-of-the-art and future microelectronic processors in energy efficiency and computational speeds by seven- and four orders of magnitude, respectively. Our research: 1) explores energy-efficient photonic neurons with graphene-based electro-optic modulators (devices theme); 2) develops ultrafast, scalable, photonic neural network architectures in silicon photonics platforms (architectures theme); and 3) advances the understanding of processors for generalized neuromorphic computing tasks including deep learning and nonlinear optimization (applications theme).
Quantum Photonic Neural Networks
Quantum photonic neural networks (QPNNs) are brain-inspired, reconfigurable quantum circuits that leverage the strengths of photonic platforms (for multiplexing, low latency, low powers) and can be trained to implement high-fidelity quantum operations. To date, proposed QPNNs assume idealized components, including a perfect 𝜋 Kerr nonlinearity. We are studying realistic QPNNs that suffer from fabrication imperfections leading to photon loss, imperfect routing, and weak nonlinearities, showing that they can learn to overcome these errors. We recently discovered that an optimal network size balances imperfections versus the ability to compensate for lacking nonlinearities. With a sub-optimal 𝜋/10 effective Kerr nonlinearity, using the example of a Bell-state analyzer, we showed that a network fabricated with current state-of-the-art processes could achieve an unconditional fidelity of 0.891, which increases to 0.999999 if it is possible to precondition success on the detection of a photon in each logical photonic qubit. Our results could enable the construction of viable, brain-inspired quantum photonic devices for emerging quantum technologies for quantum simulators, to process or reduce the noise of quantum states, and quantum machine learning.
Laser Dynamics and Networks
We are studying laser dynamics and developing photonic integrated circuits in III-V and hybrid silicon photonics platforms with networks of lasers for spike-based information processing and chaos for optical secure communications. This paradigm takes advantage of ultrafast and low-energy phenomena in photonic devices, and has the potential for orders of magnitude performance improvement over current processing systems. Spike-based encoding scheme has both analog and digital qualities, and, via careful interleaving of the two representations, is potentially capable of achieving the high performance and low power characteristics of analog systems while simultaneously being cascadable and scalable. Our approach exploits both function-oriented models in optoelectronics and integrated interconnection strategies based on fiber optical networking.