RESEARCH HIGHLIGHTS

The field of high-performance computing (HPC) has been changing rapidly over the last several years. Gone are the days where simply adding more computer cores to a machine or designing a slightly better chip will lead to large performance increases. Instead, HPC centers have begun focusing on extracting maximum efficiency from each compute node on their systems, training users to make improvements in writing and porting their scientific applications to HPC systems, and, increasingly, looking to supplement their traditional flagship machines with next-generation technologies that can lead to scientific breakthroughs.

One of those key technologies, quantum computing, promises a new way of computing that can tackle certain highly complex problems more efficiently than traditional HPC systems. But to date, quantum computers are still in their infancy, and even the most powerful quantum computers might be more computationally competitive with your cell phone than world-class supercomputers.

In the interest of scaling quantum computers to new heights, a research team led by Prof. Simon Trebst at the University of Cologne has been using the JUWELS supercomputer at the Jülich Supercomputing Centre to better understand how specific quantum mechanical phenomena can help researchers design more powerful, more accurate quantum computers that are more resilient to less-than-optimal environmental conditions.  

“Over the past four years, we have seen many quantum computing devices come online that just a decade ago we could not imagine becoming a reality in the near future,” said Trebst. “This is partially because people now accept that so-called ‘noisy, intermediate-scale’ (NISQ) devices are a valid path to pursue in addition to designing pristine, so-called ‘topological’ quantum computers. These are made by some of the world’s biggest computer companies such as IBM and Google, and their availability allows more teams to actually go out and run codes on these systems so we can evaluate what they can do.”

Until the team’s most recent compute cycle, it had been primarily using JSC resources to focus on understanding quantum materials more broadly. But in the last year, the team has pivoted to focus more on using HPC to model and investigate increasingly larger quantum circuits, paving new ways to use quantum mechanics to increase quantum computers’ capabilities.

Classical calculations light the way toward advanced quantum entanglement

As scientists began to develop technologies for observing atomic and sub-atomic interactions, they came to understand that the rules governing interactions of matter at the scale visible to the naked eye—so-called classical physics—started to break down at extremely small scales. For example, in classical physics, all matter can only exist in one state or another, meaning that charged particles can occupy only one space or appear in one specific atomic position. In quantum systems, however, particles can theoretically occupy two states simultaneously, called superposition.

Physicist Erwin Schrödinger explained this concept using a cat in a box as an analogy: Schrödinger explained that if a cat is placed in a box alongside poison set to release at the first sign of atomic decay, the cat must be considered to exist in a state of being alive and dead—rather than alive or dead—as long as the box remains closed.  For quantum physicists, especially those focused on building larger, more powerful quantum circuits for things like quantum computers, the search for these “cat” states of subatomic particles is of utmost importance. By effectively manipulating and controlling these “cat” states, researchers can extend the length of connected subatomic particles interacting with one another in proximity—also known as entanglement—and increase computational power.

The traditional method of designing a quantum computer is to start with a given state of qubits—quantum particles that are serving as the “bits” in a classical system. Then, perform a series of individual calculations, or gate operations, and finally measure the qubits’ states to extract information to get a result.  

Trebst and his collaborators look for new ways to increase entanglement in a quantum system, and they found that these moments of measuring the system can be done mid-circuit to ultimately help extend entanglement. Because researchers cannot observe these tiny particles’ ultra-fast interactions, they need to model the various ways that these particles can interact and change over time. Further, even modest numbers of qubits have extremely large numbers of possible configurations or states, meaning that modelling these systems requires researchers to simulate the myriad “trajectories” for a given quantum circuit. To that end, the team needs access to HPC resources like those at JSC.

While having chains of qubits in perfectly uniform quantum “cat” states is ideal for quantum computing, these systems must be maintained in extremely controlled environments free from even the smallest influence of temperature fluctuations, vibration, or any other minor change. The team wanted to understand the role of small imperfections on a system, such as one or two particles pointing in the wrong direction from its neighbors, to determine the threshold where these small-scale errors would degrade the accuracy of a calculation.

The team identified a structure in its trajectory calculations that can be mapped to a well-known classical physics model in statistical mechanics called the random-bond Ising model (RBIM). Four decades ago, in fact, a Japanese physicist named Hidetoshi Nishimori did his PhD thesis studying the RBIM and charting out a line where temperature fluctuations and disorder—how magnetic particles are able or not able to interact with one another based on their respective magnetic moments —would essentially cancel one another out. Nishimori found a distinct line on a graph showing this cancellation in a classical physics system, but when Trebst, his postdoc Dr. Guo-Yi Zhu and collaborators from Harvard University ran their quantum calculations, they found their system sat exactly along the same line, leading them to name the state “Nishimori’s cat,” in honor of the physicist.

“This is one of those extremely rare examples where many calculations being done on a national supercomputer, like the one in Jülich, led us to a situation where we can do a relatively short paper-and-pencil calculation to confirm it,” Trebst said. The finding has implications for how to better design “fault-tolerant” quantum circuits that remain accurate in less than perfect conditions. The team’s work is published in Physical Review Letters and has a preprint publication online reporting additional simulations together with a team at IBM Quantum.

Computing a solid foundation for next-generation technologies

Trebst indicated that the team’s work falls in line with the early state of quantum computing generally—it works on scaling these systems to be larger and more resilient. While classical systems like JUWELS use varied architectures, the methods of solving a particular type of scientific problem are similar, and researchers are easily able to reproduce one another’s work using the same computational code on a different machine. Since quantum computing is a relatively new technology, these systems are not as consistent from calculation to calculation. “This field is still in its infancy, and at this stage, many of these calculations using quantum circuits are more like guided experiments,” Trebst said. JSC is fostering the growth of quantum computing through early investments in the technology, including the Jülich UNified Infrastructure for Quantum Computing (JUNIQ) facility, which offers users access to the D-Wave Advantage quantum annealer, one of the most powerful quantum systems available today.

He went on to say that traditional HPC was supporting the team’s work by standing as a third pillar between theory and experiment. Trebst explained that quantum systems are still heavily limited by how many qubits researchers can effectively join—the depth of the quantum circuit—and that these devices are not as tolerant to imperfections as current classical machines. Trebst is excited not by the idea of quantum computers overtaking classical systems, but rather using classical systems to help improve quantum computers to the point where these two technologies can play complementary roles in an HPC center’s hardware ecosystem.

“Most of us in the research community are not interested in the perceived ‘competition’ between these two technologies, but rather to look for constructive ways to join the advantages of these two technologies and make the most use out of both. I think the idea that we will be running these systems side-by-side is feasible within the next decade,” he said.

-Eric Gedenk

Related Publication: Zu, G, N. Tantivasadakarn, A. Vishwanath, S. Trebst, and R. Verresen. “Nishimori’s Cat: Stable Long-Range Entanglement from Finite-Depth Unitaries and Weak Measurements,” Physical Review Letters 131 (2023). DOI: https://doi.org/10.1103/PhysRevLett.131.200201