Northwestern University physicists are using NVIDIA technology to tackle the computationally demanding tasks hindering quantum research.
Northwestern theoretical physicist Jens Koch and his research group develop advanced, large-scale computer simulations to model superconducting circuits for quantum computing devices. As these simulations grow in size and complexity, the mathematical calculations driving the models become prohibitively slow.
To help overcome this computational hurdle, Koch has turned to NVIDIA graphics processing units (GPUs) and software for accelerating his simulations. GPUs can run thousands of calculations simultaneously, making them ideal for handling the heavy math behind quantum simulations.
In early tests, NVIDIA GPUs accelerated one time-intensive task by 16 times compared to the speed of a standard CPU computer processor.
"NVIDIA has incredible expertise in GPUs, which make computations much faster," Koch said. "With NVIDIA's help, our algorithms will run much faster because they will use GPUs instead of regular computing processors. We have hit a major bottleneck for our simulations, so, for our work to continue to move forward, we are depending on this code."
An expert in quantum systems and quantum information processing, Koch is a professor of physics and astronomy at Northwestern's Weinberg College of Arts and Sciences. He's also deputy director of Fermilab's Superconducting Quantum Materials and Systems Center and co-director of the Northwestern-Fermilab Center for Applied Physics and Superconducting Technologies. Danyang Chen and Lambert Lin, both graduate students in Koch's research group, are supporting the new project.
The backbone of Koch's research is a supercomputing qubit. Similar to a "bit" in a classic computer, qubits are the most basic building block of information in a quantum computer. Supercomputing qubits are qubits made from tiny electrical circuits.
With the ability to exist in multiple states at once, qubits have opened the door to powerful and fast quantum computing. But designing and controlling qubits is notoriously challenging. That's why Koch and his team's work is crucial. If scientists can simulate quantum devices before building them, they can pinpoint and troubleshoot errors in advance.
"There is a shared challenge among researchers working in quantum computing," Koch said. "That challenge is that quantum mechanics is fragile. As we make these systems larger, we have trouble controlling them. It's hard to make quantum physics behave properly inside of a computer. We model these systems to help others understand how they behave and find the performance bottlenecks, so they can ultimately improve their performance."
Within the past few years, Koch's team developed an open-source package for modeling superconducting qubit systems. Called scQubits, the software package has been downloaded more than 340,000 times.
"It tries to simplify modeling for quantum devices," Koch said. "Our package makes it as streamlined and as efficient as we can with current computers."
Now, Koch and his team are integrating NVIDIA's cuQuantum software development kit to accelerate two crucial mathematical calculations underpinning their simulations. In one test, the team used GPUs on a standard task that tracks how qubits interact with one another and evolve over time. Using GPUs increased the task's speed by 16 times.
The team also is testing the GPUs on a task to compute the energy levels of large-scale supercomputing circuits, which requires intense calculations to find hidden properties in giant grids of numbers.
Although testing the GPU code within the simulations is still in its early stages, Koch and his team hope the upgrade could dramatically reduce runtimes and enable simulations that were previously impossible. The team aims to release this new version of scQubits integrated with the latest NVIDIA cuQuantum functionalities in the coming weeks. This will enable a significant improvement in the throughput for superconducting qubit design workflows.
"Scientific computing is central to quantum research," Koch said. "NVIDIA recognizes that and wants to help us make our software package run faster. We're all very excited to see how these tools can boost the quantum physics research community."
