Sign In

Communications of the ACM

ACM News

Department of Energy Accelerates Emerging Technologies

View as: Print Mobile App Share:
The $8-million project will focus on computational chemistr for the masses.

The Pacific Northwest National Laboratory aims to use computational chemistry expertise to advance the next generation of molecular modeling capabilities.

Credit: Nathan Johnson/Pacific Northwest National Laboratory

The innovation cycle for an emerging technology can take years, and sometimes decades, to move from the idea stage to a commercialized technology, a timeframe that the U.S. Department of Energy's Director of the Office of Science, Asmeret Asefaw Berhe, says the U.S. can no longer indulge if it is to stay ahead of its global competition.

The result is the agency's new $73-million program to "Accelerate the Transition from Discovery to Commercialization." The two-year program, launched with a call for proposals in February, announced in September it will fund 11 projects.

"We need to provide more paths to connect existing use-inspired basic research to applied research to shorten the time between lab bench and use. This funding will help solve challenges and gaps in the transition process," explained Berhe.

One of the major projects, funded with $8 million, is computational chemistry for the masses, directed by the Pacific Northwest National Laboratory (each two-year program will be directed by a National Lab).

In cooperation with industrial partners Microsoft and Micron, Pacific Northwest National Laboratory (PNNL) dubbed its project Transferring Exascale Computational Chemistry to Cloud Computing (TEC4), the goal of which is to create a cloud computing environment specialized to provide exascale computational chemistry as a service (CCaaS), in order to accelerate the discovery of new materials with the predictive capabilities of exascale computing.

"Oak Ridge National Lab already has an exascale supercomputer and Argonne National Labs will finish installing a second facility later this year, but there is a long waiting line to get access to these exascale computing resources," said PNNL principal investigator Karol Kowalski. "Our goal is to democratize access to exascale-performance computational chemistry tools in the cloud."

In order to achieve this goal, 30 software experts at PNNL, as well as colleagues at Central Michigan University, Lawrence Berkeley National Lab, Louisiana State University, University of Texas, and the University of Utah, will create software workflows that can be configured individually for acceleration in the cloud, resulting in overall performance rivalling that which can be achieved by exascale supercomputers.

"Computational chemistry simulates molecules at the quantum mechanical level, which creates several different bottlenecks in its workflow," said Kowalski. "We plan to orchestrate these workflows so that during execution, they switch to different accelerators that each provide exascale performance for that segment of the workflow."

The cloud architecture for this will be Microsoft's Azure complemented by accelerators provided by Micron, which can be configured for each segment of the CCaaS workflow. PNNL researchers already have been working with Microsoft Azure's cloud architecture with limited success, but with the boost from the DoE "Accelerate" program, Kowalski hopes to identify and resolve the bottlenecks with Micron hardware—in particular, its Compute Express Link (CXL), designed to be configurable to mitigate bottlenecks between computing resources and dynamic random access memory.

"Modeling molecular interactions at the quantum mechanical level requires access to lots of memory in different stages of the workflow," said Kowalski. "Micron's technology is a prerequisite for executing the novel machine learning algorithms we are designing for CCaaS."

Foremost in their CCaaS architecture is reducing the cost and increasing access to computational chemistry for new materials exploration, while at the same time reducing energy consumption by providing exascale acceleration only where and when it is required.

The first year of the program will be all about putting together the TEC4 computational workflow with the segments identified that need exascale acceleration on demand. The second year will be about putting together a turnkey CCaaS architecture and presenting working applications to the DoE.

Two TEC4 applications will be targeted for the DoE demonstration. The first will explore scalable degradation methodologies for "forever chemicals" (polyfluoroalkyl substances, abbreviated as PFAS), hazardous chemicals used to make everyday household products that are building up in the environment since they don't break down easily over time. The second demonstration application will try to identify "green" catalysis chemicals for the production of safer fertilizers.

Angstrom Era

Another national laboratory project receiving an $8-million award under the program hopes to accelerate sub-nanometer (angstrom-era) semiconductor design strategies, led by principal investigator Chang-Yong Nam at the Brookhaven National Laboratory. His proposal, titled "Angstrom Era Semiconductor Patterning Material Development Accelerator," aims to solve a bottleneck holding back the timely development of improved photoresists required for patterning semiconductors below the one-nanometer (10-angstrom) technology node with extreme ultraviolet (EUV) 13.5-nanometer-wavelength light, Nam says, pointing out that current photoresists require relatively long EUV exposure time, compared to previous-generation 193-nanometer wavelength deep ultraviolet (DUV) light.

"The longer EUV exposure time required by traditional polymer photoresists reduces productivity by as much as 60% compared to the previous-generation DUV technology," said Nam.

According to Nam, carbon, the main element in conventional organic photoresists, has little EUV sensitivity, requiring the photoresist to be exposed longer for patterning. The solution, Nam contends, is to infuse traditional organic polymer photoresists with inorganic atoms that have better EUV sensitivity, such as tin, indium, and hafnium. Conventional chemical synthesis of inorganic-containing photoresists is complex and time-consuming, which explains Nam's quest to find a better way.

"We have had success with vapor-phase infiltration of inorganic elements into organic photoresists after the spin-coating step on a wafer," said Nam, "but there are many variables that need to be studied and tuned for each candidate inorganic material, starting with how the infused inorganic element is chemically interacting with the organic matrix, and how much to add."

Nam's team will be trying to fine-tune that using machine learning to correlate the different variable values with the best EUV patterning performance for angstrom-era photolithography. One stumbling block is that EUV production machines cost $200 million—far beyond what typically can be afforded for lab work. That is why Nam and his colleagues will use inexpensive electron-beams (e-beams) to roughly approximate EUV, followed by using an EUV beam line from Brookhaven's and Lawrence Berkeley Lab's x-ray synchrotrons. The idea is to tune "proxy" variables from e-beam and synchrotron lines using machine learning to correlate them to EUV variable settings. At the end of the project, Nam and his colleagues hope to have machine-learned a look-up table that specifies how to properly tune EUV parameters from properly tuned e-beam and synchrotron beam line variables.

"We also hope to determine fundamentally 'how' inorganic elements increase photoresist sensitivity, which will ease exploring the very large space of hybrid material compositions," said Nam.

Besides Brookhaven National Labs and Lawrence Berkeley National Labs, the University of Texas, and California State University will be helping in the project.

Other awards being administered to the national laboratories include two for the Thomas Jefferson National Accelerator Facility in Virginia; "Advance Superconducting Integration Process Enabling Sustainable Hardware for AI and Quantum Computing," under principal investigator Anne-Marie Valente, and "Next Generation High-Power Compact Accelerators for Industrial Applications" under principal investigator John Vennek.

At Argonne National Laboratory, principal investigator Jerry Nolen will oversee a project titled "Tele-robotics to Tele-Autonomous Robotics for Isotope Production."

Oak Ridge National Laboratory principal investigator Venugopal Varma will oversee a project on "The Maintainable Fusion Pilot Plant."

At the SLAC National Accelerator Laboratory in Menlo Park, CA, principal investigator Simon Bare will pursue a project called "Integrated Platform to Predict Degradation of Catalysts for Sustainable Conversion of Alternate Feedstocks to Fuels and Chemicals," while principal investigator Julie Segal plans to run a project on "3D Integrated Sensing Solutions."

Finally, at Colorado's National Renewable Energy Laboratory, principal investigator Davinia Salvachua Rodriguez will oversee a project on "Real-time Sensing and Adaptive Computing to Elucidate Microenvironment-Induced Cell Heterogeneities and Accelerate Scalable Bioprocesses."

The 11 projects received $38 million in Fiscal Year 2023, which ended September 30; another $35 million in funding is slated for fiscal 2024.


R. Colin Johnson is a Kyoto Prize Fellow who ​​has worked as a technology journalist ​for two decades.


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account