Quantum Simulations Reach 200 Qubits with New Accelerated Computing Method

Researchers at OTI Lumionics Inc., led by Seyyed Mehdi Hosseini Jenab, Brandon Henderson, and Scott N. Genin, have developed a parallel implementation of the iterative qubit coupled cluster (iQCC) method, achieving a significant advance in quantum chemistry simulations. This new approach overcomes limitations in classically emulating quantum circuits by distributing computational load across multiple nodes and utilising GPU acceleration, resulting in speedups exceeding two orders of magnitude. Importantly, the iQCC method avoids the ‘barren plateau’ problem, enabling stable and efficient calculations on complex systems. Demonstrating the power of this technique, the team successfully simulated ruthenium catalysts – industrially relevant molecules – requiring between 100 and 124 qubits, surpassing the accuracy of Density Matrix Renormalization Group calculations and suggesting that the threshold for demonstrable quantum advantage in chemistry may be considerably higher than previously anticipated, potentially exceeding 200 qubits.

Genin at OTI Lumionics Inc. The approach overcomes limitations in classically emulating quantum circuits by distributing computational load across multiple nodes and utilising GPU acceleration, resulting in speedups exceeding two orders of magnitude. The iQCC method avoids the ‘barren plateau’ problem, enabling stable and efficient calculations on complex systems.

Demonstrating this capability, the team successfully simulated ruthenium catalysts – industrially relevant molecules – requiring between 100 and 124 qubits. This was enabled by a parallel, GPU-accelerated implementation of the iterative qubit coupled cluster (iQCC) method, reshaping expectations for when quantum computers will definitively outperform classical systems for chemistry problems. The significance of this achievement lies in extending the boundaries of classical computational chemistry, allowing for the study of increasingly complex molecular systems that were previously intractable.

The iQCC method efficiently distributes computational workload and avoids the ‘barren plateau’ phenomenon, a common obstacle in quantum algorithm training, ensuring stable calculations. The ‘barren plateau’ arises from the vanishing gradients encountered during the optimisation of quantum circuits, hindering the learning process. By employing an iterative approach and carefully managing the circuit structure, iQCC mitigates this issue. NVIDIA GPUs completed ground-state calculations for ruthenium catalysts – molecules vital in industrial processes – in 1.2 to 45 hours. Distributing Hamiltonian terms—representing the total energy of a molecule—across multiple compute nodes and utilising Graphics Processing Units (GPUs) facilitated complex calculations known as Pauli contractions. Pauli contractions are fundamental operations in quantum chemistry calculations, involving the contraction of multiple Pauli operators. These operations are computationally intensive, and GPUs provide significant acceleration due to their parallel processing capabilities. This technique breaks down the complex calculations – represented by the ‘transformed Hamiltonian’, a simplified map of the molecule’s energy landscape – into smaller parts and spreads these across multiple computer nodes. The transformed Hamiltonian is a key component of the iQCC method, allowing for a more efficient representation of the molecular system and reducing the computational burden.

Error rates dropped significantly, enabling more reliable and accurate simulations. Ruthenium catalysts, industrially important molecules used in various chemical processes like hydrogenation and metathesis, underwent full ground-state calculations within 1.2 to 45 hours on NVIDIA GPUs. The transformed Hamiltonian enabled efficient computation for these simulations, allowing for a more detailed and accurate representation of the electronic structure of the ruthenium catalysts. This level of detail is crucial for understanding and optimising catalytic activity.

Distributed qubit coupled cluster calculations for ruthenium catalyst simulations

The iterative qubit coupled cluster (iQCC) method tackles the exponential growth of computational demand in quantum chemistry by cleverly distributing the workload. It is a sophisticated mathematical recipe for calculating a molecule’s energy, designed for quantum systems. The coupled cluster method is a highly accurate technique for determining the electronic structure of molecules, but its computational cost scales exponentially with system size. iQCC addresses this limitation by iteratively solving the coupled cluster equations, distributing the calculations across multiple processors and GPUs.

Ruthenium catalysts were simulated using NVIDIA GPUs on systems ranging from 100 to 124 qubits. The method confines calculations to a manageable computational space, enabling accurate results with current hardware. This is achieved through a combination of efficient algorithms and parallel processing, allowing for the simulation of larger and more complex molecules than previously possible. This approach is fundamental to designing new materials and drugs, yet classical computers struggle with the exponential increase in computational demand as molecular size grows. Accurate molecular modelling is essential for predicting the properties of new materials and designing drugs with improved efficacy and reduced side effects.

No prior method matched this. Simulations of ruthenium catalysts – important industrial compounds – requiring 100 to 124 qubits were demonstrated, a feat previously thought beyond reach. These catalysts play a critical role in numerous industrial processes, including the production of pharmaceuticals, polymers, and fine chemicals. Accurate molecular modelling offers potential for designing new materials and drugs, accelerating the discovery process and reducing development costs. Some experts questioned whether achieving these simulations with 100-124 qubits truly signals a near-term path to quantum advantage, suggesting that further advancements in both quantum hardware and algorithms are needed.

However, this work recalibrates expectations for when quantum computers will outperform conventional methods. By demonstrating accurate modelling of complex ruthenium catalysts – used in industrial processes – with currently available qubit counts, the authors suggest the threshold for quantum advantage may be sharply higher, potentially exceeding 200 qubits. This implies that achieving a definitive quantum advantage in chemistry requires not only more qubits but also more sophisticated algorithms and error correction techniques. The current results suggest that the initial expectations of achieving quantum advantage with fewer than 100 qubits were overly optimistic.

Speed doubled, representing a significant improvement in computational efficiency. These simulations surpassed previous limitations using current quantum computing capabilities. A new computational approach extends the practical limit of classical simulations for complex molecular systems to 124 qubits. Combining this refined algorithm with parallel processing on graphics processing units, scientists overcame a key obstacle in quantum chemistry: the exponential growth in computational demand as molecules become larger. It avoids the ‘barren plateau’, which typically halts calculations before useful results emerge, allowing for the exploration of a wider range of molecular systems and chemical reactions.

.