5

Nanoscale Device Characterization Division

The Nanoscale Device Characterization Division is home to scientists and engineers with deep expertise in design, fabrication, and characterization of novel nano- and atom-scale engineering solid-state materials and devices. The vision of the division is to transform nano- and atom-scale technologies by advancing measurement science and fundamental knowledge. The division’s mission is to develop and advance the measurement and fundamental knowledge needed to characterize nano- and atom-scale engineered materials and solid-state devices for innovation in information processing, sensing, and future quantum technologies.

The division is organized into the following five groups: the Advanced Electronics Group, Alternative Computing Group, Atom Scale Device Group, Nanoscale Processes and Measurements Group, and Nanoscale Spectroscopy Group. The division’s strategic areas of research are as follows: theory and measurements for alternative computing; microscopies for quantifying emergent properties in quantum materials and devices; advanced microelectronics—integration, characterization, and authentication; and atom-scale solid state devices—theory, fabrication, and measurements.

TECHNICAL QUALITY OF THE WORK

Overall, the quality of the technical programs is impressive; much of the scientific work at the division ranks among the best of the world in its field of research.

Among the highlights of the division’s technical accomplishments, the extraordinary scanning probe effort is a standout. The Alternative Computing Group has developed an impressive platform to enable testing of prototype artificial intelligence hardware, available to NIST and its partners. By allowing new computing prototypes, such as cross-bar or neuromorphic chips, to be physically integrated onto a foundry-fabricated CMOS wafer, to be probed, packaged, and tested as a system, new levels of system-level testing have become available. Donor and dot platforms in the quantum area have also made great strides and are clearly positioning themselves to make important contributions.

In a forward-looking research area such as the development of future nanodevices, it can be very difficult to assess the needs of industry with respect to NIST’s mission. The division needs an improved understanding of which nano- and atom-scale engineered materials and solid-state devices are of greatest interest to industry, and what measurements and fundamental knowledge gaps exist with respect to their characterization. Division researchers are pursuing promising initiatives whose proposed directions are aligned with cutting-edge research and NIST mandates. However, in some instances, there are other research groups that are pursuing similar directions and that are ahead or in better institutional positions to succeed, by virtue of funding and/or infrastructure. The division’s researchers need to better define their current complementary or competitive role, vis a vis external research efforts, and where their research needs to be in the future, so that leadership can strategically align resources and laboratory mandates to position the researchers for success in these new initiatives. The division needs to develop clearer metrics and strategic planning to define and accomplish success and growth goals. It is important to clearly link the choice of metrics to the NIST mandates such as advancing U.S. competitiveness, being the laboratory for industry, and supporting development of measurement standards. Given the newness of this division after a recent reorganization, such gaps in metrics and strategic planning are not unexpected.

In the area of quantum computing, increased interaction with industry would beneficially inform the work being carried out at the division. The phosphorus atomic precision platform could significantly help further understanding of control limits (i.e., manufacturing inaccuracy and imprecision) and benchmarking of Hamiltonian translation to physical systems. In addition, the unique capabilities in low-temperature scanning tunneling microscopy (STM) could provide new information on two-level systems (TLS) that impact the ever-important issue of T₁, qubit energy relaxation time. While there is an extensive literature regarding TLS providing a phenomenological framework and conjectures for sources of TLS, the atomic-scale sources of energy relaxation have yet to be identified and characterized conclusively particularly for next-generation materials of interest. By combining 8 mK temperatures and 7–8 micro electron volt energy resolution, using the STM tip to identify defects (e.g., TLS) and then carry out spectroscopy on these defects could lead to unprecedented understanding and characterization of TLS.

The common thread across the division’s work in alternative computing appears to be at the device level, as embodied in the non-volatile memory crossbar/platform work. The impending end of Moore’s Law seems to imply that today’s emphasis on machine learning based on graphics processing units (GPUs) and their huge attendant energy consumption will necessarily end. With this alternative computing effort, the division is placing a bet that what will follow GPU- and accelerator-based machine learning will be similar to what came before it—biology-inspired neuromorphic, low-energy arrays of nonlinear memory elements such as memristors. Crossbar neuromorphic devices utilizing memristors of either magnetic tunnel junctions or WO₃ (tungsten oxide) are of significant interest in a number of companies. The stochastic nature of filament formation and control to vary resistance is actively being investigated and is a challenge in issues of control, but importantly an opportunity for ultralow-power artificial intelligence computing.

TECHNICAL EXPERTISE OF THE STAFF

The caliber and expertise of the senior researchers and associates in the Nanoscale Device Characterization Division are impressive.

ADEQUACY OF RESOURCES

In general, the facilities of the division are excellent. Much of the equipment is state of the art with some pieces being one of a few in the world or one of a kind. The installation of the photoemission electron microscopy (PEEM) system and planned inclusion and coupling of a ultraviolet (UV) and deep UV femtosecond light source constitutes an exciting addition to the suite of nanoscale analytical tools. PEEM can be used to investigate differences in the work functions of individual grains, for example, and the inclusion of a femtosecond source could provide information on key dynamics at the nanoscale and femtosecond time domain.

During the review, there was little mention of supercomputing (or even large-scale computing) resources available to support the division’s technical work. One theorist said he did not need much support, and another researcher said he had adequate access to computing support. Other national laboratories, such as Lawrence Berkeley National Laboratory, however, point to their onsite supercomputer as being a fundamental required resource for analytics and modeling and argue that continuing development of even more capable supercomputers is essential to scientific progress at all levels, including basic research. In that context, the division’s apparently modest computing resources appear minimal by comparison. One presenter suggested that the COVID-19 pandemic had helped his research by making more computing cycles available. “Doing the best with what you have” is a generally laudable attitude, and it is in evidence here, but under-resourcing the computing systems available to the division’s scientists could effectively limit the scope of their work or delay its completion. If more advanced and capable computer systems were to be made available, the division may benefit from adding

additional skill sets to their current mix. In particular, algorithms experts can be extremely helpful in matching what domain experts (the division’s current personnel) know to the tools and methods that best suit high-end computing systems.

During the review, insufficient time was made available for discussion of the adequacy of human resources. One apparent inadequacy might be a need to add a person with systems expertise to the team. This gap may affect the group’s ability to address issues relating to wear-out mechanisms of non-volatile memories. Accessed enough times, any given cell can no longer reliably retain state. Flash memory of the kind in USB keys have this characteristic, and the general solution (a systems level approach) is to provide a wear-leveling controller that reassigns addresses as needed, to keep any one cell or group of cells from being overused. Presentations by division researchers mentioned this problem of wear-out in passing, but there was no concerted effort to address it directly. There are both system reliability and security implications to this problem that will require higher-than-device-level research to solve.

EFFECTIVENESS OF DISSEMINATION OF OUTPUTS

The division is using well the traditional methods of output dissemination, including publications in high-quality journals, timely workshops, and direct connections to outside projects. The division is also using impressive new channels of communication. For instance, making the atomic force control software available on GitHub is a very good first step toward disseminating that work to the widest possible audience. A useful follow-up step could be to proactively contact whoever downloads that code to see if support or collaboration would be useful. Another example was the crossbar platform for artificial intelligence. The design of that system was done with an eye toward facilitating its duplication outside of NIST (i.e., using industry standards, appropriate languages, and affordable hardware). This would have the intended effect of making future collaboration much more feasible and effective.

Much of what NIST does is basic research, and fundamental facts about how nature is organized are by definition not patentable. However, the innovative work described to the panel showed ample examples of material that would clear the novelty threshold required of patents by U.S. law. Unless there are legal or policy restrictions at NIST, patenting such work seems worthy of consideration.

In the area of quantum computing, there is an increasing need for standards development. There is a need for standards relating to characterization not only of the qubits but of the elements of the qubits such as properties of quantum dots as well as donor and Josephson junction-resonator structures. For example, standards and reliable technique development for tunnel barrier characterization, charge noise, defect identification, as well as valley splitting (e.g., spatial uniformity) are all areas where groups report results from various methods without any authoritative understanding of the comparisons of accuracy and precision. Benchmarking of novel computing schemes such as analog quantum simulation is also an opportunity, given NISTs present position. Leadership in this area is needed, and perhaps a NIST-led workshop series might be timely to help the community to converge in some of these areas.

GENERAL CONCLUSIONS AND RECOMMENDATIONS

Because NIST is tasked with developing new characterization technologies, U.S. companies would benefit greatly from the capabilities within NIST to develop next-generation tools for addressing a range of technical challenges at the nanoscale for companies without the resources to develop these analytical probes themselves.

Adequate computing resources, including supercomputers where appropriate, are essential for conducting research at the advanced levels evident at NIST. Such resources may not be currently available to the division’s researchers.