Si-Integrated BaTiO₃—Emergent Platform for Integrated Optical Computing

Traditional computing based on CMOS technology is nearing physical limits in terms of miniaturization, speed, and power consumption. Consequently, alternative approaches are under investigation. The most promising is based on a “brain-like” or <i>neuromorphic computation </i>scheme; another is quantum computing. Both of these approaches can be realized optically using silicon photonics (SiPh), in which the fundamental unit is an efficient, ultra-low power broadband optical modulator. A complete or partial switch from electrons to photons in computing would be revolutionary. Such technology ultimately requires the integration of both active and passive photonic elements on a single chip. As silicon modulators suffer from relatively high-power consumption and large size, materials other than silicon are now being considered for realizing these compact energy-efficient modulators. I will discuss our recent progress in integrating ferroelectric perovskite BaTiO₃(BTO) with SiPh for the purpose of fabricating modulators that exploit the linear electro-optic effect, as well as other passive optic elements. As-grown Si-integrated BTO demonstrates low optical loss on the order of 1 dB/cm that is sufficient for passive elements. We also show microring resonators fabricated in the monolithic Si-integrated BTO system with 50 mm radius that demonstrate very large Q-factors (Q &gt; 5 × 10⁴) confirming exceptionally low loss. Mach-Zehnder modulators based on BTO have demonstrated V_πL as low as 0.23 Vcm (0.014 Vcm in plasmonic devices), modulation bandwidth of 100 GHz, and data rates above 250 Gb/s. In addition, Si-integrated BTO-based devices have shown robust cryogenic performance. Importantly, unlike LiNbO₃, BTO is fully compatible with Si foundry processing. These properties will enable neuromorphic circuit architectures that exploit shifting computational machine learning paradigms, while leveraging current semiconductor manufacturing infrastructure. This will result in a new generation of computers that consume less power and possess a larger bandwidth. A particularly elegant implementation of optical neuromorphic computing is a reservoir computer, a computational framework suited for temporal/sequential stream data processing. Such a computing system consists of a reservoir for mapping inputs onto a high-dimensional space and a readout for pattern analysis from the high-dimensional states in the reservoir. The reservoir is fixed and only the readout is trained with a simple method such as linear regression and classification. The major advantage of reservoir computing compared to other neural networks is its fast learning, which substantially reduces training cost. I will discuss our progress in implementing an integrated photonics reservoir.