Energy-efficient data centers and high-performance computing enabled by on-chip printed microelectronic optical component and circuit manufacturing

Printed electronics offer the potential to achieve low-energy consumption and integration of novel optoelectronic nanomaterials onto silicon wafers for manufacturing photonic integrated circuits (e.g., silicon photonics). However, there are limited choices in commercial semiconductor inks, and most of them do not apply to optoelectronic applications. In this Phase I project, Iris Light Technologies and Argonne National Lab (PI: Yuepeng Zheng) will overcome the electronic conductivity limitations of existing printed semiconductors to enable improved generation efficiency of on-chip light emitters. Improved ink synthesis and printing process reproducibility and yield will be demonstrated. The Phase I deliverables are: (1) an improved semiconductor ink with 20% larger conductivity, and (2) prototype LEDs emitting at the wavelength relevant to cloud data center communications, e.g. 1310 nm. Iris Light will lead the ink development, device design, and opto-electronic component characterization, while ANL will lead the printing technique and process advancement for precise ink deposition and patterning. In Phase 2, the team will integrate the prototype devices directly onto foundry chips to demonstrate foundry-compatible manufacturing of onchip embedded lasers and light-emitting diodes (LEDs). A major outcome of the program is the technology transition to a state-of-the-art 300 mm semiconductor foundry located within the US with a mission to grow domestic manufacturing capability for the silicon photonics market (AIM Photonics). Iris Light has worked with this foundry for over four years towards on-chip component development including gratings, optical filters, polarization control, photodetectors, and more. The advances achieved in this program will foster increased US manufacturing competitiveness in the emerging space of printable microelectronics and next-generation optoelectronic materials. Photonic chips (photonic integrated circuits - PICs) have gained significant traction for their high-performance and low-cost components in the optical communications industry (datacom/telecom) valued at over $1 billion today and growing at 45% annually. The growth of cloud infrastructure has fueled demand for scalable, denser integration of onchip optics for pluggable optics that have reached bandwidths exceeding 100 Gb/s. From an energy perspective, the data center market already accounts for ~2% of global emissions, 10% of electricity consumption, and is expanding at an unsustainable rate. The rise of high-performance computing (HPC) presents an additional challenge. In the Fall of 2021, Argonne National Lab (ANL) retired its third-generation petascale machine, Mira, with the announcement of transitioning to the exascale using a device called Aurora which consumes high levels of electricity (60 MW/year), enough to power 42,000 homes. Increasing data transfer efficiency (more data per unit of energy) contributes to decarbonization and is complementary to efforts focused on shifting energy production to sustainable sources. Both approaches are vital to our function as a society in the long run. The efficiencies enabled by photonic chips are driving network bandwidths into the terabitper- second regime and simultaneously decreasing data energy cost per bit. Silicon photonics is already commercially deployed and poised to decrease the need for high-capacity cooling systems, a major driver in power consumption in data centers. While there has been immense progress in silicon photonics, the search for on-chip laser sources continues despite decades of research. To overcome the ‘silicon laser problem’, Iris Light Technologies is developing a new class of on-chip lasers based on photonic ink gain material printed on silicon wafers. These new on-chip lasers will not only drastically reduce the size of circuits, but also enable higher-bandwidth network interconnections between compute and memory nodes as required by energy-efficient disaggregated networks coming online.