Accurate and Efficient Quantum Computations of Molecular Properties Using Daubechies Wavelet Molecular Orbitals: A Benchmark Study against Experimental Data

Cheng-Lin Hong, Ting Tsai, Jyh‐Pin Chou, Peng-Jen Chen, Pei-Kai Tsai, Yu-Cheng Chen, En-Jui Kuo, David J. Srolovitz, Alice Hu, Yuan‐Chung Cheng, Hsi‐Sheng GoanPRX 2022 · DOI

TL;DR

This paper introduces a novel approach using Daubechies wavelet molecular orbitals to significantly improve the accuracy and efficiency of quantum computations for molecular properties. By employing this method, researchers can achieve results comparable to high-level classical calculations with a minimal number of qubits, making accurate quantum chemistry simulations more feasible on near-term quantum hardware.

What Problem Does This Paper Address?

Quantum computation holds immense promise for revolutionizing quantum chemistry by potentially overcoming the exponential scaling of classical methods. However, current quantum algorithms for molecular simulations are largely limited to small molecules due to the significant number of qubits and complex quantum circuits required. This limitation stems from the need to represent molecular systems accurately, often necessitating large basis sets that translate to high computational costs. Existing minimal basis sets, like STO-3G, are computationally inexpensive but lack the accuracy needed for reliable predictions. The challenge, therefore, is to find a way to represent molecular Hamiltonians accurately and efficiently, reducing the quantum resource requirements without sacrificing predictive power.

Methods

The core of this work lies in the adoption and application of Daubechies wavelet functions as a basis set for constructing molecular orbitals. Unlike traditional atomic orbitals, Daubechies wavelets offer several advantages:

  • Localization: They are localized in both real and reciprocal space, which is beneficial for computational efficiency and preconditioning.
  • Completeness: The Daubechies wavelet basis set is complete, eliminating superposition errors inherent in incomplete basis sets like STO-3G.
  • Adaptivity: They allow for a multiresolution approach, using fine grids for chemically relevant regions and coarser grids elsewhere, leading to efficient representations.

The authors construct minimal basis sets of Daubechies wavelet molecular orbitals (DW MOs) by performing Hartree-Fock (HF) self-consistent-field (SCF) calculations using the BigDFT code. These DW MOs are then used to build a second-quantized Hamiltonian. For quantum computation, this Hamiltonian is transformed into qubit operations using parity encoding and exploiting Z2 symmetry to reduce qubit count. The Variational Quantum Eigensolver (VQE) algorithm is employed to calculate molecular energies. To benchmark their approach, they compare results from DW/UCCSD (Daubechies wavelet with Unitary Coupled Cluster Singles and Doubles ansatz) and STO-3G/UCCSD with classical Full Configuration Interaction (FCI) calculations and experimental data. They also explore the impact of noise on quantum simulations using a heuristic hardware-efficient circuit ansatz.

Datasets

The paper does not introduce new datasets. Instead, it utilizes:

  • Experimental data for vibrational frequencies and bond lengths of H2, LiH, and H2O for benchmarking.
  • Classical computational results from FCI calculations using various standard Gaussian basis sets (6-31G, cc-pVDZ, cc-pVTZ) as a reference for accuracy.
  • Simulated noisy quantum computer data from the IBM Qiskit simulator, using noise parameters from the "ibmq-santiago" device.

Key Insights and Ideas

  • Daubechies wavelets offer a superior minimal basis: The study demonstrates that a minimal basis set constructed from Daubechies wavelets provides a significantly more accurate representation of molecular Hamiltonians compared to the traditional STO-3G minimal basis.
  • Accuracy without increased qubit count: The key breakthrough is achieving high accuracy with a minimal number of spin orbitals (and thus qubits), comparable to much larger, high-level Gaussian basis sets, but at a computational cost equivalent to STO-3G.
  • Vibrational frequency as a sensitive benchmark: The paper highlights that vibrational frequencies are more sensitive indicators of computational method quality than bond lengths, especially for benchmarking.
  • Quantum computing cannot compensate for a poor basis: The results clearly show that even with advanced quantum algorithms like UCCSD, a low-quality basis set (STO-3G) leads to inaccurate predictions, underscoring the critical importance of the basis set choice.
  • Heuristic ansatzes are viable for NISQ devices: The study suggests that low-depth heuristic ansatzes, when combined with a high-quality Daubechies wavelet Hamiltonian, can yield accurate results even under noisy quantum conditions, making them suitable for near-term quantum computers.

Key Results

  • H2, LiH, and BeH2: DW/UCCSD calculations for H2, LiH, and BeH2 yielded vibrational frequencies in excellent agreement with experimental data, comparable to or better than high-level FCI calculations using large Gaussian basis sets. For H2, the DW/UCCSD vibrational frequency error was 1.53%, while STO-3G/UCCSD was 12.0%.
  • Water molecule (H2O): The DW/UCCSD method accurately predicted the bond length and bond angle of water, with the bond angle prediction being even better than cc-pVDZ/FCI. The number of spin orbitals used was 12, significantly less than the 48 used in cc-pVDZ/FCI.
  • Computational Cost: The DW/UCCSD method achieved this accuracy using a number of spin orbitals comparable to the STO-3G basis set, implying a similar computational cost for the classical part of the calculation and a reduced qubit requirement for the quantum part.
  • Noise Resilience: Even with simulated noise, the DW/UCCSD approach, particularly with heuristic ansatzes, showed promising results, with error mitigation techniques bringing the results into excellent agreement with exact FCI calculations.

Limitations and Future Work

  • Limited to Small Molecules: The current study focuses on small molecules (H2, LiH, BeH2, H2O). While the method shows promise, its scalability to larger, more complex molecules needs further investigation.
  • Heuristic Ansatzes and UCCSD: While heuristic ansatzes are shown to be effective for NISQ devices, the paper also benchmarks against the UCCSD ansatz, which, despite its accuracy, requires a much deeper circuit and is thus less suitable for current hardware. The trade-off between ansatz complexity and accuracy in the context of DW MOs could be further explored.
  • Noise Mitigation: While error mitigation techniques were applied, the inherent limitations of NISQ devices mean that noise remains a factor. Further research into more robust error mitigation strategies tailored for DW MO-based calculations could be beneficial.
  • Systematic Improvement: The paper mentions that the Daubechies wavelet basis offers means for systematic improvement through finer grids. Exploring this systematic improvement for larger systems and more complex properties would be a natural next step.

Summary updated

Abstract

The molecular orbitals for H_2, LiH, and water are calculated using a quantum algorithm based on the Daubechies wavelet theory: Accurate results are achieved, while keeping the number of qubits minimal.
BibTeX
@inproceedings{hong2022accurate,
  author = {Cheng-Lin Hong and Ting Tsai and Jyh‐Pin Chou and Peng-Jen Chen and Pei-Kai Tsai and Yu-Cheng Chen and En-Jui Kuo and David J. Srolovitz and Alice Hu and Yuan‐Chung Cheng and Hsi‐Sheng Goan},
  title = {{Accurate and Efficient Quantum Computations of Molecular Properties Using Daubechies Wavelet Molecular Orbitals: A Benchmark Study against Experimental Data}},
  booktitle = {PRX},
  year = {2022},
  doi = {10.1103/prxquantum.3.020360},
}
Figure