NERSC Pursues Growing Opportunities in Quantum Information Science

Building world-changing quantum information technologies for the scientific community and beyond requires interdisciplinary research, and as a leader in team science, Lawrence Berkeley National Laboratory (Berkeley Lab) is blazing a trail. Across the Lab, multi-faceted quantum information science (QIS) research laboratories, centers, and projects are partnering with industry and academia to fabricate and test quantum-based devices, develop software and algorithms, build a prototype computer and network, and apply these innovations to enable  breakthroughs in multiple science areas.

Recognizing that some combination of quantum and classical computing will be key to the application and adoption of quantum information resources going forward, the National Energy Research Scientific Computing Center (NERSC) at Berkeley Lab is preparing for the day QIS-enabled computing will be available to its community. In addition to growing staff expertise and collaborations in this area, NERSC recently established the QIS@Perlmutter program and issued a call for proposals. The goal of the program is to allow a broad set of projects to harness the power of Perlmutter to develop QIS devices and techniques, and for NERSC to gain a better understanding of scientist’s use cases and their classical and quantum requirements. So far, NERSC has awarded more than 250,000 Perlmutter GPU node hours to 16 QIS projects that cover a wide range of topic areas.

Learning how best to integrate classical and quantum computing capabilities is top of mind these days, according to Katie Klymko, a staff member in NERSC’s Advanced Technologies Group who previously worked in Berkeley Lab’s Applied Math and Computational Research Division and is now spearheading the QIS@Perlmutter program. In this Q&A, Klymko discusses the importance of bridging the gap between classical computing and quantum computing for applications in chemistry, physics, materials science, drug discovery, and more, and how NERSC’s role in QIS will be that of a centralized resource for these efforts.

What inspired NERSC to launch its Quantum Information Science (QIS) program?

As a currently classical supercomputing facility, we understand that at some point incorporating quantum hardware into our resources will be essential, and it is clear that QIS technology will be extremely useful to our users. There remain many unanswered questions however, and we are working to figure out how quantum hardware can be combined with high performance computing to meet our users’ needs. Will different quantum computing technologies have different requirements for coupling to classical computing resources? Will the quantum hardware sit right next to the classical supercomputing equipment, or will it be in a different facility altogether? No one really knows, and in some sense that is the research focus right now.

It was these kinds of questions that inspired us to launch the QIS@Perlmutter program. The response to the QIS@Perlmutter call for proposals was fantastic and included research teams from industry, academia, and national labs who are working on applications in materials science, optimization, chemistry, and applied math (among other topics). For example, Rigetti Computing, a superconducting qubit company, is studying hybrid quantum-classical optimization algorithms at scale. Another project is being led by a current Berkeley Lab postdoc, Lindsay Bassman, who will be using Perlmutter’s GPUs in combination with NVIDIA’s cuQuantum SDK to simulate disorder-induced quantum phase transitions.

We are excited that so many research teams want to test their QIS concepts on Perlmutter. “Quantum” has become such a buzzword, but how useful quantum technologies will be, what the timeline for their development and implementation is, and which applications QIS actually impacts is not really clear from what you read in the news. Quantum hardware is basically a big physics experiment, and there are different types of technologies that have different benefits. We have to figure out how to pass data back and forth between the quantum and classical units, and that is definitely one of our research focuses at this point.

Why does NERSC think QIS will be important for its users?

Many of the science problems NERSC users are focused on are quantum mechanical in nature. In fact, solving quantum mechanical problems on a classical computer has been a big part of NERSC’s workload for decades. That doesn’t mean all of them will be better on quantum hardware, but it is likely that quantum hardware can help address these research challenges.

That said, scale is currently an issue because the quantum hardware that is available is quite small. People are trying to scale it up, but we’re not quite at the point where we can do anything that we couldn’t do on a classical computer. You can still run something on a quantum computer and do some verification on a classical computer that the calculation was correct because it’s still small enough. But as the quantum hardware scales up and you have large amounts of data coming from that hardware, you won’t necessarily be able to check the calculations using classical resources.

For example, NERSC is collaborating with the Advanced Quantum Testbed (AQT) at Berkeley Lab on an experimental project leveraging HPC with AQT’s superconducting quantum processors. It’s going to be at a small scale because that is what we can do right now, but hopefully it will give us a better understanding of the exchange of resources and help us set up a pipeline for future projects that want to incorporate both AQT’s state-of-the-art superconducting quantum platform and NERSC. Such experiments will give us valuable insights into when NERSC should provide quantum computing resources for its users and in what form.

Has NERSC been seeing an increase in the amount of quantum computing-related research it supports?

NERSC’s users have been working on QIS-related research for many years; one of the more recent notable examples was in 2017 with a 45-qubit simulation on Cori. We are now seeing an explosion of interest, as evidenced by the response to the QIS@Perlmutter program. Based on the proposals we received, the number of projects specifically focused on quantum algorithms for quantum applications (as opposed to classical algorithms for quantum applications) is increasing and will continue to increase. And as quantum hardware begins to scale up, I expect to see quantum computing algorithm development as a large component of NERSC’s workload.

What science applications will benefit from QIS technologies in the near future?

I think that one of the first true demonstrations of quantum advantage will be for chemistry, such as modeling large-scale correlated molecules. A lot of this type of work is currently done on NERSC’s classical supercomputers, but we can’t do it as well as we’d like to with classical methods because the problems are just too hard. Ideal quantum hardware will allow us to model large, strongly correlated molecular systems accurately, impacting our understanding and prediction of chemical reaction pathways. Other areas where quantum hardware could be beneficial include optimization and machine learning – areas that many NERSC users are already actively involved in.

Right now we are trying to figure out where exactly QIS will have the most impact for our users, and on what time scale. We are working to demonstrate quantum advantage within both hardware development and algorithm development.

What are some emerging trends in the development of quantum algorithms?

Coming up with algorithms that reduce the number of qubits and gates required for near-term algorithms is definitely a trend. Hybrid algorithms – where part of the work is off-loaded to classical resources, which allows for shorter quantum circuits – are one approach, and Berkeley Lab is a leader in the development of hybrid algorithms devoted to coupling quantum and classical computers.

Another exciting development is the realization that many of the major quantum algorithms (notably quantum search, quantum phase estimation, and Hamiltonian simulation) employ a particular technique known as the quantum singular value transformation (see Gily´en, Su, Low, and Wiebe, ACM STOC 2019 and Martyn, Rossi, Tan, Chuang, PRX Quantum, 2021). Recognition of this algorithmic primitive as a key component in many currently important quantum algorithms has provided a more unified understanding of the landscape of quantum algorithms and could generate future developments. One example could be studying whether hybrid algorithms can be constructed, and potentially enhanced, by incorporating QSVT-based subroutines.

What does all of this mean for the future of science?

Depending on who you talk to, we are anywhere from 5 to 20+ years away from a fully error-corrected quantum computer. Once that is available, we can do things like study molecules at a scale we haven’t been able to even imagine classically, which will allow us to make predictions for chemical reaction pathways and enable advances in catalysis design, drug development, and many other areas. Such a quantum computer will impact many research efforts in the chemical, physical, and materials sciences.

I’m really excited to be working on quantum information science as a part of the Berkeley Lab environment. We have all the necessary pieces to tackle large-scale problems, including experts in algorithm design, hardware development, HPC, chemistry, and materials science. This puts us in a great position to tackle meaningful problems and demonstrate true quantum advantage. Quantum hardware will augment, not replace, our classical resources and will allow us to tackle many problems that are currently intractable. It will also change the way we perform computational research, and the next generation of HPC centers will likely look completely different from what they are today. NERSC sees its role as helping the R&D community fit all of the pieces together as we create the QIS ecosystem.

NERSC is a U.S. Department of Energy Office of Science user facility.