PhD Thesis Defense

Zoom link: https://caltech.zoom.us/j/82251869328

Integrated circuits, notably chips made in CMOS, have been the primary driving force behind decades of technological advancement. Combined with fiber-optic communications, this foundation has led to growth in the field of integrated photonics. The combination of electronics and photonics on the same chip has been envisioned for as long as photonics has existed, but it has lacked a robust and scalable solution in bulk CMOS, which accounts for over 90% of all chips made today.

This thesis discusses efforts to scale monolithic photonics in bulk CMOS to a point where it can be made in any process without process modifications, and devices are reliable enough to make complex systems with up to tens of thousands of components. We start by presenting the first monolithic optical clock distribution network in bulk CMOS, where all photonic, optoelectronic, and electronic circuits are integrated in the same foundry-fabricated chip. This offers a potentially low-power, low-skew, and low-crosstalk alternative to traditional clock distribution networks, which can account for as much as 50% of a processor's total power consumption. Along the way, waveguiding is demonstrated in the most advanced process to date, breaking a previously thought limit and indicating its potential for scaling. The same platform then serves as the basis for the first monolithic IQ coherent receiver in bulk CMOS, enabling applications in LiDAR and biology. After a brief detour to silicon photonics, where digital beamforming is demonstrated for the first time in optical phased arrays, we describe a new method of implementing optical phased array receivers. By breaking a long-standing assumption that fields must co-propagate for their interference to be detected, an optical phased array receiver without any optical antennas, phase shifters, or even waveguides is presented. This approach yields a 2D array with orders of magnitude improvement in element density, and the theory introduced provides insight into other applications that may be affected by the result.