QIS@Perlmutter Fuels Discoveries and Collaborations in Quantum Computing

Since its launch in 2022, NERSC's QIS@Perlmutter program supporting novel quantum information science (QIS) projects has exceeded expectations, generating strong engagement among researchers and fueling numerous discoveries in quantum computing. Initiated as a resource for users to investigate the interplay between classical and quantum computing using the power of the Perlmutter supercomputer, the program has met those goals and additionally yielded prolific research results and an extensive list of scientific publications. Thus far, QIS@Perlmutter has awarded hundreds of thousands of GPU hours to 27 research teams, leading to new algorithms, hardware analysis, and insights into quantum computing as well as collaborations that will help both users and NERSC advance this promising field of computing.

“We believe that quantum computing has the potential to be a game-changing technology for our users,” said NERSC Senior Scientific Advisor Richard Gerber. “So we started this program to look at how quantum computers could be used to support our users’ scientific research and how our facility could adapt to deliver this resource.”

With time allocated from the NERSC Director’s Reserve, the QIS@Perlmutter program continues to attract new proposals as it heads into its third allocation year. NERSC Advanced Technologies scientists Katie Klymko and Daan Camps, who lead the QIS@Perlmutter program, have been amazed by its productivity.

“QIS@Perlmutter has surprised us with the amount of interest there is,” said Klymko. “We didn’t really know how many QIS researchers would need classical HPC resources for their work.”

In its first few years, the program has drawn significant interest, including by users who haven’t worked with NERSC before, and the first few allocation years have yielded exciting work with both scientific users and industry partners.

Supporting the Future of Quantum with Perlmutter

One QIS@Perlmutter research theme is focused on understanding and reducing the noise inherent to today’s quantum computers, which can affect their functioning. A team led by Ang Li at Pacific Northwest National Lab 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 characterized the noise. In another QIS@Perlmutter research project, Li and his team ran high-performance numerical simulations of deep quantum circuits to verify the accuracy of low-energy nuclear physics applications efficiently. The approach employed several novel methods for accelerating the numerical simulation, an important method for verifying the quantum circuits used to simulate low-energy nuclear states. These simulations ordinarily place demanding memory and processing requirements on conventional simulation methods.

“Perlmutter provides nice support of the high-speed GPU interconnect for us to overcome the performance bottleneck of the simulator,” said Li. “Ready access to Perlmutter has been very helpful, because we are developing quantum software tools, and we want users to try them and use our tools on the system.”

Another user who’s published extensively through QIS@Perlmutter access, Liang Tan, a scientist at Berkeley Lab’s Molecular Foundry, has used the program to look closely at materials modeling. Tan’s research is focused on creating ideal quantum defects for quantum information applications and studying how to make them brighter and more effective.

“A lot of these simulations are very computationally expensive, and this is where Perlmutter comes in,” Tan said. “What we need to do is to solve the Schrödinger equation on a large number of electrons – these electrons define the quantum state of the material. These simulations contain many atoms and many electrons and Perlmutter is ideal to study systems of this size.”

“Perlmutter is a very big machine and it's also got very advanced hardware, and both of these things are essential,” he added. “We need the calculations to run fast, but we also need to run a lot of them.”

An Unexpected Industry Connection

Camps and Klymko have also been pleased with the amount of industry participation in the program, even though it is something they actively solicited, targeting companies in the quantum computing space. “We want to develop good relationships with industry,” said Camps. “By providing the growing quantum computing industry with world-class computational resources, we aim to make an impact and accelerate the development of quantum technologies.”

Industry derives value from relationships with NERSC, and the entire scientific community benefits when works are published and tools are released for general use. For example, Xanadu, a Toronto-based quantum computing company, has developed algorithms on Perlmutter that have since been integrated into their open-source quantum computing software framework, PennyLane. Since collaborating with NERSC through the QIS@Perlmutter program, Xanadu has also joined the NERSC Science Acceleration Program (NESAP), a collaborative effort in which NERSC partners with code teams, vendors, and library and tools developers to prepare for advanced architectures and new systems.

Lee James O’Riordan, a senior quantum software developer at Xanadu, said the QIS@Perlmutter program allocation was key to refining the circuit-cutting algorithm that makes their PennyLane software particularly effective. The method divides quantum circuits into smaller sub-circuits to make processing on today’s quantum computers possible, and then stitches the results back together afterwards. “We are focused on enabling whatever type of workloads a user is interested in,” said O’Riordan. “That means targeting both simulated quantum devices as well as real-world quantum devices. So for our model, it's been great that we can run everything on the GPUs, and some of the larger-scale workloads we were running through PennyLane were only possible through the Perlmutter system.”

Klymko points to Xanadu’s work at QIS@Perlmutter as a notable development. "If a user wants to simulate a really large system and doesn’t necessarily have the resources to simulate the whole thing, Xanadu came up with a way to simulate smaller pieces and then stitch them together,” she said. “It gives people the option to simulate large-scale systems.”

Developing Software for Simulation

Among all QIS@Perlmutter users, software development has certainly been a main research theme, said Klymko. “Running simulations at scale is not trivial, and we have multiple users from this program who can do this now,” she said. “They’re using Perlmutter to simulate really large-scale systems that are pushing the boundaries of what you can do classically, and they’re developing software to simulate these large-scale systems.” She adds that there has been some very interesting recent experimentation among users in augmenting quantum computing with machine learning.

“Simulations are really the bulk of our projects,” said Camps. “Users are either developing their own simulation software, testing at scale, or using existing simulators to run applications they’re interested in.”

Over the course of the three years of the QIS@Perlmutter program, Camps said it has been particularly interesting to see it get easier to run large-scale simulations—he points to industry user NVIDIA and the quantum software toolkit, which includes the cuQuantum simulator and CUDA-Q platform for hybrid quantum-classical computing that the company has deployed at NERSC. “It’s really maturing and they’ve been giving trainings to other users,” he said. “We’re watching the development of a community of quantum users with the know-how about QIS.”

Growing collaborations is key to the future of quantum computing at NERSC, and as the QIS@Perlmutter program moves into another allocation year, NERSC staff are looking forward to more growth. “It's really a great example of using today's technologies to help advance the state of the art to enable new technologies for users’ science,” said Gerber.

The National Energy Research Scientific Computing Center (NERSC) is a U.S. Department of Energy Office of Science User Facility that serves as the primary high performance computing center for scientific research sponsored by the Office of Science. Located at Lawrence Berkeley National Laboratory, NERSC serves almost 10,000 scientists at national laboratories and universities researching a wide range of problems in climate, fusion energy, materials science, physics, chemistry, computational biology, and other disciplines. Berkeley Lab is a DOE national laboratory located in Berkeley, California. It conducts unclassified scientific research and is managed by the University of California for the U.S. Department of Energy.