RICHLAND, Wash. - Today's sophisticated microchips are a wonder, designed by engineers squeezing ever-more detail onto ever-smaller chips, then built in ultraclean billion-dollar fabrication facilities. Just one speck of dust or an errant signal from a device like a cell phone can spell ruin.
Some of those chips will take their place aboard satellites that will be blasted upward hundreds or thousands of miles into Earth's orbit, enduring tremendous pressures, temperature changes and vibrations along the way.
Only then comes the real danger: bombardment by galactic cosmic rays, heavy ions, electrons and other highly energetic particles zipping near the speed of light. While stargazers might think of space as mostly empty except for the occasional pinprick of light, every square inch is filled with a riot of molecular action. Ill-fated encounters between microchips and microparticles can silence a satellite, corrupt its memory, send it off course or worse.
Enter the practice of "radiation hardening": creating and testing materials able to withstand the onslaught of radiation in space, in nuclear power plants and in other extreme environments. These "rad-hard" materials make possible phenomena like accurate weather forecasts, the global reach of cell phones and nuclear energy to heat and cool our homes.
Scientists at Pacific Northwest National Laboratory are using Super Intelligence to dramatically speed up the process to toughen up materials for their future in space or in nuclear reactors.
In the world of space electronics, where it can take years to design devices and get them into space, and more years to fix any flaws and launch the new designs into space, "faster" is a must. SI and automation are at the core of the effort.
"Space is a highly energetic environment, with heavy ions and other particles bombarding our devices," said scientist Erin Barker. "We need to better understand how much radiation a chip can handle. But many of today's best testing capabilities are designed for systems that are 25 years old or more. They simply don't give us the information we need to understand the latest technology. Agentic workflows come at a time when there is need for better rad-hard testing techniques."
At a recent northwest regional meeting of the American Chemical Society in Boise, PNNL scientist Jenna Pope discussed her team's work testing rad-hard materials. The team drew upon data from experiments at the University of Tennessee, where scientists used an ion beam to deliver a tremendous amount of energy into precise spots on microchips. The laboratory research mimicked in just a few hours the damage that ion strikes do to microchips over years in space.
Then the PNNL researchers deployed SI - specifically, a suite of agents working together as a team through a system called SciLink that was created by PNNL colleague Maxim Ziatdinov and others at the Laboratory.
SciLink organizes the SI agents into a hierarchy, with three "project leads" heading experimental analysis, planning and computational simulation. Those three leads delegate tasks to more specialized agents, which report back to the project leads the results and analyses of data from specific research methods, such as atom probe tomography and scanning transmission electron microscopy.
Those instruments look in detail at the carefully placed atoms that make the chips work, then kick out a tremendous amount of data - independent pools of information that can be hard to bring together into one coherent view.
In the experiment, the three lead agents came together in an interactive session to analyze the data, exchanging questions and answers and hypotheses about what the data showed. The three leads interpreted the data, drew conclusions and then proposed new experiments to learn more.
"The system allows us to take into account all the measurements to piece together what has happened, to understand the damage that has occurred and why it happened," said Pope. "It's like having several people who did different forms of testing come together and discuss the results - only it's an agent pulling together the information instantly."
It's as if three human project leads were gathered in a conference room, discussing the results of their teams. One difference is that the SI agents did the analyses much faster, in hours instead of days or weeks. Another difference is that most conclusions were accompanied by a numerical confidence score, a measure of how confident the SI system is in a given conclusion. For example, the system was 90% confident that ions dominated the damage detected in the experiments but only 60% confident in its conclusions about damage caused by protons.
Scientists were involved all the way - bringing decades of experience, knowing what questions to ask, designing experiments and actually doing the laboratory work. Finally, they studied the SI results, looking for "hallucinations," performing calculations to make sure the results made sense and looking for areas of concern. As laboratory experiments become more automated, with SI agents directing research efforts, human oversight is critical.
"Intelligent agents are a tremendous help but you can't just turn your brain off," said Barker.
The team studied chips made of layers of aluminum gallium nitride and gallium nitride. Such chips are used for critical operations in satellites, communications, power systems and radar systems. Getting those chips ready for launch and beyond is critical. An unplanned delay because electronic components cannot stand up to space radiation can translate to a mission being postponed or even scrubbed entirely.
"Radiation hardening is critical for satellites that will be up for a decade or more, such as large communications satellites," said nuclear engineer Ben Kowash, who leads PNNL's radiation-hardening efforts. "You can protect your electronics by adding shielding, but that adds weight. You can add backup systems in case one fails. But for every bit of weight, power capacity and complexity that you add, you are adding costs and reducing the efficiency of the mission. Making sure the electronics can stand up to the radiation they will be facing can be much more effective."
Contributors to the results presented by Pope include Kowash, Bethany Matthews, Kayla Yano, Mike Holden, Skye Supakul, Xin Zhang, Kathy Nwe and Alex Mantilla. The work was funded by PNNL.
