System Bits: July 3

CMU prof gets a shot at new supercomputer
The National Energy Research Scientific Computing Center will greet its Perlmutter supercomputing system in early 2020. The Cray-designed machine will be capable of 100 million billion floating operations per second. Zachary Ulissi of Carnegie Mellon University will be among the first researchers to use the supercomputer.

“When this machine comes online, it will be one of the largest open-science machines in the U.S.,” said Ulissi, an assistant professor in chemical engineering. “Our day-to-day work uses machine learning methods and high throughput calculations, but in the past, these tasks often had to be done in separate locations due to limited computational resources. The Perlmutter supercomputer will greatly accelerate both the data generation and the machine learning model development, allowing us to compute many more iterations of our models much, much faster.”

According to NERSC, Perlmutter is the first supercomputing system designed to enable both data analysis and simulation. Participants in this first round are encouraged to explore applications of the Perlmutter’s capabilities in three ways: simulation of complex physical phenomena, real-time data analytics through the supercomputer’s GPU architecture and cutting-edge machine, and deep learning solutions.

Ulissi and his research team will be using Perlmutter’s expanded computing power to accelerate their search for new materials that can serve as active catalysts for renewable energy chemistries. His project was chosen as part of the NERSC Early Science and Application program, and as such, the team will work with high-performance computing experts at NERSC to develop and tune GPU-accelerated machine learning methods for this new machine. The project will then be used to demonstrate to future users the impact of the new machine, and to verify that it runs according to its targeted specifications.

“We do hundreds of expensive calculations every day to search for catalysts that can more efficiently split water to produce renewable hydrogen, directly convert waste CO2 into a valuable feedstock chemical and improve the efficiency of hydrogen fuel cell vehicles,” Ulissi said. “All of these technologies are important in an increasingly electrified chemical economy.”

With current computing technology, it takes considerable time and money to analyze each catalyst in search of which will provide fuel cells with greater efficiency and capacity. With supercomputing technology like Perlmutter at their disposal, Ulissi and his team will be able to perform more of these calculations much faster, enabling them to develop technologies that will bring us closer to a zero-emissions transportation sector.

Team touts gold as quantum computing material
A joint team of scientists at the University of California, Riverside, and the Massachusetts Institute of Technology is getting closer to confirming the existence of an exotic quantum particle called Majorana fermion, crucial for fault-tolerant quantum computing — the kind of quantum computing that addresses errors during its operation.

Quantum computing uses quantum phenomena to perform computations. Majorana fermions exist at the boundary of special superconductors called topological superconductors, which have a superconducting gap in their interiors and harbor Majorana fermions outside, at their boundaries. Majorana fermions are one of the most sought-after objects in quantum physics because they are their own antiparticles, they can split the quantum state of an electron in half, and they follow different statistics compared to electrons. Though many have claimed to have identified them, scientists still cannot confirm their exotic quantum nature.

The UCR-MIT team overcame the challenge by developing a new heterostructure material system, based on gold, that could be potentially used to demonstrate the existence and quantum nature of Majorana fermions. Heterostructure materials are made up of layers of drastically dissimilar materials that, together, show completely different functionalities when compared to their individual layers.

“It is highly nontrivial to find a material system that is naturally a topological superconductor,” said Peng Wei, an assistant professor of physics and astronomy and a condensed matter experimentalist, who co-led the study, appearing in Physical Review Letters, with Jagadeesh Moodera and Patrick Lee of MIT. “A material needs to satisfy several stringent conditions to become a topological superconductor.”

The Majorana fermion, considered to be half of an electron, is predicted to be found at the ends of a topological superconductor nanowire. Interestingly, two Majorana fermions can combine with each other to make up one electron, allowing the quantum states of the electron to be stored nonlocally — an advantage for fault-tolerant quantum computing.

In 2012, MIT theorists, led by Lee, predicted that heterostructures of gold can become a topological superconductor under strict conditions. Experiments done by the UCR-MIT team have achieved all the needed conditions for heterostructures of gold.

“Achieving such heterostructure is highly demanding because several material physics challenges needed to be addressed first,” said Wei, a UCR alum who returned to campus in 2016 from MIT.

Wei explained that the research paper shows superconductivity, magnetism, and electrons’ spin-orbit coupling can co-exist in gold — a difficult challenge to meet — and be manually mixed with other materials through heterostructures.

“Superconductivity and magnetism ordinarily do not coexist in the same material,” he said. Gold is not a superconductor, he added, and neither are the electron states on its surface.

“Our paper shows for the first time that superconductivity can be brought to the surface states of gold, requiring new physics,” he said. “We show that it is possible to make the surface state of gold a superconductor, which has never been shown before.”

The research paper also shows the electron density of superconductivity in the surface states of gold can be tuned.

“This is important for future manipulation of Majorana fermions, required for better quantum computing,” Wei said. “Also, the surface state of gold is a two-dimensional system that is naturally scalable, meaning it allows the building of Majorana fermion circuits.”

Besides Wei, Moodera, and Lee, the research team also includes Sujit Manna and Marius Eich of MIT.

The research was funded by the John Templeton Foundation, the Office of Naval Research, the National Science Foundation, and the U.S. Department of Energy.

Robotic arm packs boxes with AI
Using artificial intelligence to control a robotic arm provides a more efficient way to pack boxes.

“We can achieve low-cost, automated solutions that are easily deployable. The key is to make minimal but effective hardware choices and focus on robust algorithms and software,” says the study’s senior author Kostas Bekris, an associate professor in the computer science department at Rutgers University-New Brunswick.

Bekris, Abdeslam Boularias, and Jingjin Yu, both assistant professors of computer science, formed a team to deal with multiple aspects of the robot packing problem in an integrated way through hardware, 3D perception, and robust motion.

The study coincides with the growing trend of deploying robots to perform logistics, retail, and warehouse tasks. Advances in robotics are accelerating at an unprecedented pace due to machine learning algorithms that allow for continuous experiments.

Tightly packing products picked from an unorganized pile remains largely a manual task, even though it is critical to warehouse efficiency. Automating such tasks would save companies time and money.

The study focused on placing objects from a bin into a small shipping box and tightly arranging them. This is a more difficult task for a robot compared with just picking up an object and dropping it into a box.

The researchers developed software and algorithms for their robotic arm. They used visual data and a simple suction cup, which doubles as a finger for pushing objects. The resulting system can topple objects to get a desirable surface for grabbing them. Furthermore, it uses sensor data to pull objects toward a targeted area and push objects together. During these operations, it uses real-time monitoring to detect and avoid potential failures.

Since the study focused on packing cube-shaped objects, a next step would be to explore packing objects of different shapes and sizes. Another step would be to explore automatic learning by the robotic system after it’s given a specific task.

The researchers presented the study at the IEEE International Conference on Robotics and Automation. Support for the work came from research contracts and grants from JD.com’s Silicon Valley research center and the National Science Foundation.