Perlmutter Results Show Progress in Quantum Information Science

Quantum computing is coming. Though predicted decades ago and only now in its early stages, this potentially game-changing technology and the hardware and software that support it are progressing rapidly – thanks in part to dedicated initiatives like the QIS@Perlmutter project at Lawrence Berkeley National Laboratory (Berkeley Lab).

QIS@Perlmutter aims to support research in the space of quantum information science (QIS) conducted on the Perlmutter supercomputer at the National Energy Research Scientific Computing Center (NERSC), including quantum simulation of materials and chemical systems, algorithms for compilation of quantum circuits, error mitigation for quantum computing, and development of hybrid quantum/classical algorithms.

The project kicked off in early 2022, with computing resources on Perlmutter granted to 16 research teams, and its first phase will come to an end this month. But the hard work of the first year has already begun to bear fruit in the form of early science results and growing collaborations across the field.

“We’ve been seeing exciting results come out of it in all different areas,” said Daan Camps, an HPC engineer working on the initiative, adding that program participants have so far put out 10 published results and preprints, with more to come.

“We’re very happy to see that Perlmutter has been so useful to research in QIS,” said Richard Gerber, NERSC HPC department head and senior science advisor. “The goal of the program is to give researchers access to a world-class resource like Perlmutter to help develop QIS devices and techniques for the advancement of scientific research. We think there is huge potential for quantum computation, and we’re trying to help get this new technology productively into the hands of scientists as quickly as possible.”

Early results

Some of the early results of QIS@Perlmutter include the following:

• Researchers at Pacific Northwest National Laboratory (PNNL), the University of Toronto, and UC Santa Barbara made progress in understanding and reducing the noise inherent to today’s quantum computers, which can throw off their functioning. A team led by Ang Li at PNNL developed a Bayesian approach to identifying parameters in standard quantum noise models, then tested it on standard quantum algorithms and showed that the model accurately characterizes the noise. Another manuscript from Li characterizes the structure of measured bit strings (quantum data) and uses the characterization to develop a method for quantum error mitigation.
• Lee James O’Riordan at the Toronto-based quantum computing company Xanadu debuted a method of circuit-cutting, dividing quantum circuits into smaller sub-circuits that can fit more easily on the small quantum devices that are currently available. A separate post-processing step stitches the sub-circuits’ results back together. The algorithms the team developed and tested on Perlmutter have now been integrated into PennyLane, Xanadu’s open-source quantum computing software framework.
• Researchers at Rigetti Computing performed large-scale simulations on Perlmutter to address basic questions about the entanglement structure of the quantum state of the Quantum Approximate Optimization Algorithm (QAOA), a popular method for solving combinatorial optimization problems on near-term quantum computers. Their simulations revealed that the QAOA algorithm typically has to cross a phase of high entanglement as it evolves from the unentangled initial state to the low-entanglement solution. Since quantum states with a high degree of entanglement are challenging to emulate for certain widely used classical algorithms, this research provides evidence of the classical difficulty of the QAOA algorithm.
• Kwangmin Yu at Brookhaven National Laboratory (BNL) led a team focusing on scalable quantum search algorithms, particularly the efficient execution of a standard algorithm known as Grover’s search, and their implementation on real, noisy quantum computers. Additionally, the team is working to develop efficient processes on classical computers that might work with quantum computers on large decision-making problems.
• A team led by Roel Van Beeumen developed QCLAB++, a C++ code for creating and simulating quantum circuits, a key approach for developing quantum algorithms. QCLAB++ uses the state vector approach for quantum simulation where the amplitudes of the quantum wave function are tracked and updated as if they evolve on a noiseless quantum computer. The QCLAB++ code was ported to the GPU and benchmarked on Perlmutter.

Strengthening the connection between HPC and quantum

While the initiative has offered academic researchers computing power for their work, it has also helped forge partnerships with industry. “We’ve had exploratory conversations with all types of quantum companies in the industry and have set them up with access to Perlmutter through this program,” said Camps. “We’re trying to engage more of the quantum industry with the HPC field and let them use our resources.”

Some of those industry connections have proven to be two-way streets. NVIDIA developed some of the software used on Perlmutter and used feedback from researchers in the program to fine-tune their product. “Some of this has been feeding back to NVIDIA,” said Katie Klymko, another HPC engineer working on the project. “We are able to give them comments from our users, and they have made adjustments based on that.” According to Camps, industry connections have specifically noted that the initiative has been a great resource for their own work.

In October 2022, QIS@NERSC staff hosted the first annual Quantum for Science Day, bringing stakeholders from around the field of quantum computing together virtually to discuss the current state of quantum research and what its future might look like. More than 100 unique participants from across academia, research labs, and industry logged on to learn and share information, from a morning panel discussion on quantum and HPC to hands-on tutorials on the basics of running simulations on GPUs.

The planners of the event called it a success and a place for learning, as well as a harbinger of things to come, as connections across the field continue to produce results and the role of HPC in quantum computing becomes clear.

“We definitely learned that there is a pretty big place for us to help people advance their research, companies and national labs and academia alike,” said Klymko. “There is demand for HPC resources to try and tackle these problems. We were sort of testing the waters: Do people really care about doing things at scale? And it seems like this is a time when they do.”