Symmetry-allowed circuits reduce data loading for quantum algorithms

Researchers at PsiQuantum have developed a new circuit that reduces the cost of loading initial states for quantum algorithms using matrix product states. The work demonstrates a reduction of approximately 14 to 54 times the number of Toffoli gates needed for molecular benchmarks, including simulations of low- and high-spin cytochrome P450, at 15-bit rotation precision.

This circuit prepares states directly from spin-adapted reduced tensors, avoiding explicit loading of dense data and achieving a volume reduction, Toffoli gates multiplied by qubits, of 33 to 140 times. However, the team’s extrapolations suggest that for larger systems, the process of loading these states may ultimately limit the speed of quantum simulations, prompting a need to refine both state compression and fidelity.

U(1) × SU(2) Symmetry Reduces MPS Data Loading

Householder-LKS synthesis forms the core of a new circuit that directly prepares U(1) × SU(2)-symmetric matrix product states (MPS) from spin-adapted reduced tensors, circumventing the need to explicitly load corresponding unfolded dense tensors. This approach focuses on efficiently preparing reduced tensor data, which the researchers identified as the dominant cost factor in the process. The circuit uses Clebsch-Gordan coefficients to reconstruct magnetic degrees of freedom, streamlining the data handling required for quantum algorithms.

Benchmarking against 4×4 and 4×6 Fermi-Hubbard states, alongside active-space models of both low- and high-spin cytochrome P450, revealed the practical implications of this method. Analysis of the DMRG data showed that even with a truncation threshold of 10-8, the reduced tensors themselves remain largely dense, suggesting limited opportunities for further sparsity gains. This finding highlights the importance of efficient data loading, as the volume of data, not just its sparsity, becomes a limiting factor for larger systems.

The team’s work raises a critical question regarding the scalability of DMRG-state loading in relation to downstream computational costs, such as phase estimation. They note that for systems requiring bond dimensions in the thousands or tens of thousands, the loading process could ultimately exceed the computational demands of tasks like ground-state energy estimation, which typically require 105, 106 Toffoli gates.

“This raises the question of whether the loading costs are justified,” the researchers write, highlighting a potential bottleneck in current quantum simulation workflows. While additional symmetries could further reduce data requirements, the availability of supporting DMRG software tools remains a practical consideration for implementation.

Householder-LKS Synthesis for Symmetry-Adapted MPS

Householder-LKS synthesis uses Clebsch-Gordan coefficients to streamline the reconstruction of magnetic degrees of freedom within quantum simulations, bypassing the need to explicitly load dense tensor data. This approach focuses on preparing each symmetry-allowed reduced block, a method that demonstrably lowers computational demands compared to traditional techniques. The reduced-tensor data loading itself represents the dominant cost within the process, prompting exploration of further optimization strategies. Specifically, the team achieved a volume reduction of up to 140 times when compared to dense MPS loading, a metric combining gate count and qubit requirements.

These improvements were observed at a 15-bit rotation precision, suggesting a practical pathway toward more efficient molecular simulations. While the Householder-LKS method shows promise, the researchers acknowledge that alternative unitary synthesis methods could yield further gains. They note that techniques described in prior work are orthogonal to their approach and could be combined for even greater speedups.

Clebsch-Gordan Coefficients Recover Magnetic Degrees of Freedom

By directly applying Clebsch-Gordan coefficients, the new circuit recovers magnetic degrees of freedom without explicitly loading the dense tensors typically required for quantum simulations, a process that significantly impacts computational efficiency. The resulting simplification avoids redundant calculations inherent in traditional methods of representing quantum states. The circuit prepares each symmetry-allowed reduced block and applies Clebsch–Gordan coefficients to recover the magnetic degrees of freedom, avoiding explicit loading of the corresponding unfolded dense tensors.

Benchmark Results: 4×4 and 4×6 Fermi-Hubbard States

Benchmarking against both 4×4 and 4×6 Fermi-Hubbard states revealed a significant reduction in quantum resource requirements when using the newly developed circuit. The benefits extend to molecular systems as well, with low- and high-spin cytochrome P450 serving as a key benchmark. At 15-bit rotation precision, the molecular benchmarks show ~ 14× to 54× Toffoli reductions, ~ 20× to 87× active volume reduction, and ~ 33× to 140× volume (Toffoli × qubits) reduction at the QROM-optimal point, relative to dense MPS loading.

The circuit’s design allows for competitive qubit counts alongside substantial reductions in quantum resources needed for simulations. Analysis of the Fermi–Hubbard model showed that the reduction in cost is largely driven by the Householder-LKS synthesis, though qubit overhead remains a factor.

The team’s results indicate that the approach is most effective when applied to systems with a high degree of symmetry, and that the benefits increase with bond dimension and decrease with tensor density. Despite these successes, the researchers caution that scaling to larger systems presents challenges. While the current work demonstrates a clear advantage over traditional dense MPS loading, extrapolation to larger lattices suggests that DMRG-state loading may exceed downstream phase-estimation costs, motivating co-design of state fidelity, compression and the initial-state ansatz.

Molecular Benchmarks: Cytochrome P450 Performance Gains

Molecular benchmarks reveal substantial gains in quantum simulation efficiency using a newly developed circuit for preparing spin-adapted matrix product states. This decrease in gate count is coupled with a 20 to 87% reduction in active volume, the amount of quantum resources actively used during computation, and a combined volume reduction of 33 to 140% at the point of optimal quantum read-out memory (QROM) usage.

The researchers specifically benchmarked the circuit’s performance against both 4×4 and 4×6 Fermi-Hubbard models and active-space models of cytochrome P450, an important molecule involved in drug metabolism. The observed improvements in efficiency are linked to the circuit’s ability to use inherent symmetries within the molecular systems, resulting in a 20% sparsity structure with the addition of a single U(1) symmetry, and further increasing to 5% sparsity with a second U(1) symmetry.

This targeted approach minimizes irrelevant states during calculations, focusing computational power on the most pertinent configurations. The Fermi-Hubbard speedups demonstrated an inverse relationship with tensor density, indicating that systems with lower density benefit most from this method, while maintaining a modest qubit overhead. This finding challenges the conventional wisdom surrounding initial-state preparation for quantum simulations and suggests a need for a co-design approach. “Ultimately, no modality is likely to be a winner in all scenarios,” the study notes, advocating for a diversified strategy in the pursuit of efficient quantum simulations.

Toffoli, Active Volume, and Qubit Count Reductions

The circuit achieves a 33 to 140 times decrease in the product of Toffoli gates and qubits used, at the qubit-optimal point. These gains are not limited to molecular systems; Fermi-Hubbard model simulations also exhibited speedups that increased with bond dimension and correlated inversely with tensor density, though with a modest increase in qubit overhead. However, extrapolating these improvements to larger systems reveals a potential limitation.

The researchers found that for increasingly complex lattices, the cost of loading the initial DMRG-state may eventually exceed the computational expense of the downstream phase-estimation process. This suggests a need to reconsider current approaches to initial state preparation and compression, and to actively co-design these elements alongside state fidelity and the initial ansatz itself.

Efficient Unitary Synthesis and Alternative Initial States

Employing more efficient unitary synthesis methods offers a pathway to further reduce gate counts, potentially achieving a low qubit and gate overhead for quantum simulations. The research demonstrates that protocols described in existing literature are orthogonal to their work and could be integrated for additional speedup, though the current implementation uses an older synthesis method. The scale of data loading remains a significant factor. Simulations of systems with bond dimensions of 4,000 require nearly a billion data points, even with optimistic assumptions about sparsity.

The team notes that achieving ambitious goals necessitates exploring alternative methods for compressing matrix product states or considering alternatives to DMRG-based MPS altogether, as the number of data points directly correlates with the potential Toffoli gate count without qubit-gate tradeoffs. Alternative initial state loading modalities, such as Hartree-Fock and its variants, could also offer cost benefits for lattice systems, providing additional avenues for optimization.

The analysis indicates that more performant unitary synthesis methods, like those referenced, could multiply the reduction in gate cost while maintaining qubit counts, given similar qubit-gate tradeoff ratios. However, the researchers acknowledge that the current Householder-LKS synthesis method does not achieve a general result for their reduced tensor isometries, which have arbitrary dimensions, suggesting a need for continued development in this area.

👉 More information
🗞 Quantum Loading of Charge and Spin-Adapted MPS
✍️ Zheting Jin (Ryan Levy); Modjtaba Shokrian Zini (PsiQuantum, Palo Alto, CA, USA); Matthew R. Hennefarth, Lesley Gover, Carys Harvey, Ryan Levy and Sukin Sim
🧠 ArXiv: http://arxiv.org/abs/2611.19728v1

Stay current

See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.

Avatar of Rusty Flint

Rusty Flint

Rusty is a quantum science nerd. He's been into academic science all his life, but spent his formative years doing less academic things. Now he turns his attention to write about his passion, the quantum realm. He loves all things Quantum Physics especially. Rusty likes the more esoteric side of Quantum Computing and the Quantum world. Everything from Quantum Entanglement to Quantum Physics. Rusty thinks that we are in the 1950s quantum equivalent of the classical computing world. While other quantum journalists focus on IBM's latest chip or which startup just raised $50 million, Rusty's over here writing 3,000-word deep dives on whether quantum entanglement might explain why you sometimes think about someone right before they text you. (Spoiler: it doesn't, but the exploration is fascinating)

Latest Posts by Rusty Flint: