ÁùºÏ²ÊÖ±²¥¿ª½±

Q: Is SCE something that can be applied to all market segments or are there only a select few that can take advantages of the benefits you described? What are the limitations and challenges that limit its application?

I would like to answer your question in an indirect way. Every technology is capable of doing many things. But, it ends up being used in what is considered ¡°optimal applications¡± or ¡°sweet spots¡± for that technology. This is no different for SCE. SCE is certainly capable of doing many of the things that CMOS can do but, it is not necessarily the proper use of the technology. Also, within SCE there are many families that vary in speed and power consumption. One of them in particular, adiabatic quantum flux parametron (AQFP) approaches the Landauer thermodynamic limit of power consumption making it suitable for reverse computing. Others such as efficient rapid single flux quantum (ERSFQ) excels in ultra-high-speed communications, and so on. Today SCE circuits are used in D-Wave annealing type of quantum computers deployed commercially. Given that SCE circuits will have to be cooled to 4 K, don¡¯t expect to see them anytime soon in your laptop or your smart phone regardless of their power advantage. Also, SCE circuits do not scale physically with the same ease that we enjoyed in CMOS.

Q: ÁùºÏ²ÊÖ±²¥¿ª½± was awarded a contract to develop tools for SCE under the Supertools Project. What can you tell us about this project and the progress that has been made since the project started?

IARPA selected ÁùºÏ²ÊÖ±²¥¿ª½± over three years ago to develop EDA tools for the design automation of SCE circuits, supporting two families of circuits, ERSFQ and AQFP. This effort is unusual in its scope as it also includes TCAD and Spice model development. It starts at Ab Initio physics and materials based TCAD modeling all the way to GDSII. That¡¯s a great endeavor. We finished phase 1 (2 years duration) of this effort with flying colors. Also, as a ¡°down payment¡± on phase 2, demonstrating capabilities and methodology, we delivered a synthesized, placed and routed 4-bit microcontroller, the famous AMD AM2901. That was the first fully automated circuit carried from RTL to GDSII (but without full timing, for full disclosure).  It is a demanding project with very tight schedules and acceptance criteria, but it is a showcase of ÁùºÏ²ÊÖ±²¥¿ª½±-wide collaboration among its business units (BUs). Every business unit is contributing with full engagement.

Q: Describe the primary factors and considerations that could limit the growth of quantum computing and the kinds of applications that could be targeted for the initial uses of quantum computing technology given its early stages of development?

Quantum computers are expensive and have to operate at few mK. Qubits are fickle and have a low decoherence time. Thus, error correction is a major problem towards scaling the number of qubits needed for a powerful quantum computer.

Quantum computing is powerful but is not a low power proposition. In an earlier answer I referred to the concept of the ¡°sweet spot¡± for each technology. Quantum computing is capable of solving many problems, but it is not necessarily better or more cost efficient than GPU or TPU based servers. However, there are areas that are natural candidates for quantum computers such as the simulation of synthesized molecules and enzymes, simulations that will take years with a classical supercomputer and just hours with a quantum computer. Similarly, encryption and decryption, that are almost impossible to hack, are another natural application for quantum computers.

Q: Where can people go for more information on these technologies and ÁùºÏ²ÊÖ±²¥¿ª½±¡¯ role in enablement?

Interested readers can refer to our press release which describes our participation in the Supertools program at the following . They can also download our whitepaper on the subject that was accepted by the 2020 GOMAC conference and is linked here.