Ian A. Young and Dmitri E. Nikonov
Components Research, Technology & Manufacturing Group
Intel Corp., Hillsboro, Oregon, USA
ESSDERC/ESSCIRC Leuven Sept.12-14, 2017
Abstract: An analysis of research in quantum nanoelectronics and nanomagnetics for beyond CMOS devices is presented. Some device proposals and demonstrations are reviewed. Based on that, trends in this field are identified. Principles for development of competitive computing technologies are formulated. Results of beyond-CMOS circuit benchmarking are reviewed.
TABLE I: Voltage Limitations For Computation Variables
Principle 1: Beyond-CMOS circuits require CMOS as an integral part. They will work alongside and augment CMOS computing blocks.
Principle 2: Some devices utilize collective states; this confers advantages of non-volatility or more energy efficient operation.
Principle 3: The choice for an optimal beyond-CMOS device will be determined by compatibility with an efficient and effective interconnect.
Principle 4: Low voltage devices – most direct way to low energy operation.
Principle 5: Start benchmarking with bottom up modeling of devices, build up from simple to more complicated circuits.
Principle 6: Majority gates (if easily implemented in a certain technology) enable more efficient circuits, especially for more complex computation functions.
Principle 7: Use electrical interconnects for longer propagation spans.
Principle 8: To convince the wider community, a non- volatile computing paradigm needs to be general enough to prove that it is valid for more than one architecture; while it needs to be specific enough to dispel claims that an essential aspect is missed.
Principle 9: Neuromorphic computing can be done more efficiently with beyond-CMOS circuits.
Conclusions: In summary, we have presented our view on the recent trends in quantum nanoelectronics and nanomagnetics for beyond CMOS devices, and outlined a few principles to make them realize practical computing technologies. We can pose a question for ourselves: What are the most promising directions of research? Where to double down on the effort? Among many equally important thrusts, our subjective preference is for magnetoelectric switching and neuromorphic beyond-CMOS circuits.
