NERSCPowering Scientific Discovery for 50 Years

NERSC Awards 250K GPU Node Hours on Perlmutter to 16 QIS Projects

January 21, 2022

Following a request for proposals issued in November 2021, NERSC has awarded a total of 250,000 Perlmutter GPU node hours to 16 quantum information science (QIS) projects. The awards were made available through the NERSC QIS@Perlmutter program, with time allocated from the NERSC Director’s Reserve. The goal of the program is to allow researchers to use Perlmutter to help develop QIS devices and techniques for the advancement of science.

All areas of quantum information science were encouraged to apply, including quantum simulation of materials and chemical systems, algorithms for compilation of quantum circuits, error mitigation for quantum computing, and development and testing of hybrid quantum-classical algorithms.

Awards were made to researchers in fields ranging from materials science, chemistry, and computer science to high energy physics, machine learning, applied mathematics, and condensed matter physics. Nine are from national laboratories, four from industry, and one from academia.

Here are the QIS projects awarded GPU node hours on Perlmutter:

Quantum Computing for Materials Science: Simulation of Defects in Materials for Quantum Information Science

  • Principal Investigator: Marco Govoni, Argonne National Laboratory
  • Science Area: Materials Science
  • GPU Node Hours: 25K
  • Point defects in semiconductors are promising candidates for qubits and quantum sensors. In order to realize this technology, large-scale quantum simulations of correlated electronic states are needed to understand the optoelectronic properties of these materials.

Scalable Noisy Quantum Circuit Simulation through NWQSim

  • Principal Investigator: Ang Li, Pacific Northwest National Laboratory
  • Co-Investigators: NVIDIA’s cuQuantum and NVSHMEM teams
  • Science Areas: Computer Science, Chemistry
  • GPU Node Hours: 20K
  • Simulations of quantum programs in classical HPC systems are essential for validating quantum results, understanding the effects of noise, and designing robust quantum algorithms. The goal of this project is to use Perlmutter to analyze the fidelity of quantum circuit execution for algorithms for quantum chemistry and quantum error correction.

‘Divide and Conquer’ Approach to Machine Learning-Based Decoders for the Surface Code

  • Principal Investigator: Ritajit Majumdar, Indian Statistical Institute
  • Science Area: Computer Science
  • GPU Node Hours: 25K
  • Surface codes are a family of quantum error correcting codes expected to be the model of error correction for designing a noiseless quantum computer. Better decoders are needed for large-scale application of surface codes. In particular, machine learning based decoders learn the error probability of the circuit, have a linear decoding time, and thus outperform many decoders in performance.

Maximum Likelihood Estimation of Parameterized Quantum Noise Models

  • Principal Investigator: Vincent R. Pascuzzi, Brookhaven National Laboratory
  • Science Area: Computer Science (error correction)
  • GPU Node Hours: 12.5K
  • Models for quantum noise are necessary for NISQ devices (which don’t have full error correction) to achieve quantum advantage. This project aims to use quantum data and maximum likelihood estimation to enhance these models, supporting hardware and software co-design.

Surrogate Models for Variational Quantum Algorithms

  • Principal Investigator: Wim Lavrijsen, Lawrence Berkeley National Laboratory
  • Co-Investigators: Juliane Müller, Ed Younis, and Costin Iancu, Lawrence Berkeley National Laboratory
  • Science Area: Computer Science, Optimization
  • GPU Node Hours: 25K
  • Variational quantum algorithms (VQAs) combine classical optimizers with quantum hardware to achieve an overall optimization goal. In a VQA, the QPU evaluates a costly objective function, while the classical side updates the optimization parameters. Surrogate methods, computationally cheap approximations of the costly objective function, can be used to guide the selection of points in the optimization search.

Quantum Circuit Synthesis via Large-Scale Randomized Optimizations

  • Principal Investigator: Yu-Hang Tang, Lawrence Berkeley National Laboratory
  • Science Area: Computer Science
  • GPU Node Hours: 5K
  • Robust, scalable, and approximate circuit synthesis is an indispensable step in the quantum workflow, requiring large-scale classical computations.

Large-scale Hybrid Quantum Tasking and Simulation with PennyLane

  • Principal Investigator: Lee J. O’Riordan, Xanadu Quantum Technologies Inc.
  • Co-Investigators: Sean Oh, Xanadu Quantum Technologies Inc.
  • Science Area: Computer Science, Machine Learning
  • GPU Node Hours: 25K
  • Workflows that incorporate quantum and classical components are becoming increasingly common. These workflows can be supported on Xanadu’s open-source Pennylane software that, with Perlmutter’s resources, will allow users to run large-scale hybrid-quantum computations effectively, benchmark performance on quantum hardware and simulators, and investigate the possibility of demonstrating quantum advantage in machine learning applications.

Quantum Deep Learning for High Energy Physics Data Analysis

  • Principal Investigator: Shinjae Yoo, Brookhaven National Laboratory
  • Co-Investigators: Prof. Sau Lan Wu, University of Wisconsin-Madison
  • Science Area: High Energy Physics, Machine Learning
  • GPU Node Hours: 2.5K
  • There is hope that quantum machine learning could outperform classical machine learning in classification power by exploiting a large number of qubits. One way to work towards this goal is by uniting high-energy physics analysis techniques with quantum computing advances.

Implementation of Large Qubitization Iterates on a Tensor Network Quantum Simulator

  • Principal Investigator: Nathan Fitzpatrick, Cambridge Quantum Computing
  • Science Area: Computer Science, Chemistry
  • GPU Node Hours: 25K
  • Qubitization is one of the leading approaches for quantum advantage in Hamiltonian simulation, but it has so far been difficult to test computationally because of the large number of qubits required. Perlmutter’s GPUs combined with NVIDIA’s cuTensorNet SDK will enable preparing and testing these fault tolerant circuit primitives.

The Entanglement Barrier in the Quantum Approximate Optimization Algorithm

  • Principal Investigator: Matthew Reagor, Rigetti Computing
  • Co-Investigators: Maxime Dupont, Lawrence Berkeley National Laboratory
  • Science Area: Optimization
  • GPU Node Hours: 25K
  • The Quantum Approximate Optimization Algorithm (QAOA) is a hybrid quantum-classical algorithm seeking to solve combinatorial optimization problems. Investigating the role of entanglement in QAOA on fixed lattices of physical qubits, as for solid state qubit fabrics, will shed light on the limit of classical simulations and pave the way towards quantum advantage.

Benchmarking QCLAB++ on Perlmutter GPUs

  • Principal Investigator: Roel van Beeumen, Lawrence Berkeley National Laboratory
  • Science Area: Applied Mathematics, Computer Science
  • GPU Node Hours: 12.5K
  • GPU-enabled high performance quantum linear algebra computations are necessary for efficient quantum compilation of quantum circuits.

Simulating Boson Localization with Quantum Computers

  • Principal Investigator: Lindsay Bassman, Lawrence Berkeley National Laboratory
  • Science Area: Condensed Matter Physics, Materials Science
  • GPU Node Hours: 5K
  • Understanding disorder-induced phase transitions is important for utilizing cold atoms and superconductors in quantum hardware. Computationally studying quantum phases of matter requires large-scale quantum simulators and HPC resources.

Large-Scale Model-Based Optimization by Quantum Monte Carlo Integration

  • Principal Investigator: Kwangmin Yu, Brookhaven National Laboratory
  • Science Area: Optimization
  • GPU Node Hours: 25K
  • Quantum computing holds the promise of fundamentally altering the computing landscape for applying optimization techniques to address large scale problems of practical interest, which are considered intractable on even the world’s fastest classical supercomputers. Developing distributed and hybrid quantum-classical algorithms for optimal decision making that can scale to problems involving thousands of decision variables requires distributed and GPU-accelerated computing nodes.

Quantum-Inspired Approaches for Full Configuration Interaction

  • Principal Investigator: Robert M. Parrish, QC Ware Corporation
  • Co-Investigators: Sam Stanwyk, NVIDIA
  • Science Area: Computer Science, Chemistry
  • GPU Node Hours: 25K
  • New quantum algorithms are needed for large-scale quantum chemistry calculations. Perlmutter will enable computer scientists to develop and simulate how these algorithms will work on quantum computers.

Optimization and Scalability of Tensor Network Quantum Simulator QTensor on GPUs

  • Principal Investigator: Yuri Alexeev, Argonne National Laboratory
  • Science Area: Computer Science
  • GPU Node Hours: 1K
  • To explore the quantum advantage of the QAOA algorithm over classical solvers, we need to simulate large quantum circuits on supercomputers. To achieve this goal, we are developing a tensor network simulator, QTensor, using cuTensorNet in a close collaboration with NVIDIA engineers. One of the major challenges in the development is optimization and scalability of QTensor to run efficiently on GPU supercomputers. In particular, it is important to address the load balancing between CPU and GPU to achieve the best performance. We will use Perlmutter to study the scalability of QTensor.

Evaluation of Quantum Authentication Schemas to Establish User-Centered Security Solutions

  • Principal Investigator: Sanchari Das, University of Denver
  • Science Area: Computer Security, Authentication Protocols
  • GPU Node Hours: 1K
  • Quantum Authentication (QA) is an emerging concept in computer security that creates robust authentication for organizations, systems, and individuals by applying the laws of quantum physics. We plan to implement a few QA protocols to study, analyze, and determine the user side of these protocols in the real world.

Generative Quantum Eigensolver

  • Principal Investigator: Alan Aspuru-Guzik, University of Toronto
  • Science Area: Computer Science, Machine Learning, Chemistry
  • GPU Node Hours: 20K
  • This project aims to revolutionize the ground state search in computational chemistry and condensed matter physics through the development of the Generative Quantum Eigensolver (GQE).

About NERSC and Berkeley Lab
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. »Learn more about computing sciences at Berkeley Lab.