Universal 2-Local Symmetry-Preserving Quantum Neural Networks for Fermionic Systems

Ge Yan Member, IEEE, Kaisen Pan, Ruocheng Wang, Mengfei Ran, Hongxu Chen, Junchi Yan Senior Member, IEEE · 2024

Read Universal 2-Local Symmetry-Preserving Quantum Neural Networks for Fermionic Systems by Ge Yan Member, IEEE, Kaisen Pan, Ruocheng Wang, Mengfei Ran, Hongxu Chen, Junchi Yan Senior Member, IEEE (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-04825.txt.

Introduction

Ge Yan and five co-authors design a quantum neural network ansatz for fermionic systems that keeps the symmetries of physics in place while it stays fit to build on real hardware. Variational quantum algorithms hunt for ground states by tuning circuits with parameters. Circuits that skip symmetries waste effort, while ones that heed them often need high-order operations that no lab can run.

Their Hamming Weight Preserving ansatz uses just two-local operators. The team prove which terms are a must and which are enough for those operators to span the whole symmetry subspace. That proof upends the view that complex interactions are a must. Tests on computers confirmed exact approximation of arbitrary unitaries inside the subspace. The same ansatz went to work, with no change, on distinct fermionic models, which shows how wide its use can be.

Catalog record

CREATORGe Yan Member, IEEE, Kaisen Pan, Ruocheng Wang, Mengfei Ran, Hongxu Chen, Junchi Yan Senior Member, IEEE
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE71 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-04825

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Universal 2-Local Symmetry-Preserving Quantum Neural Networks for Fermionic Systems Ge Yan Member, IEEE, Kaisen Pan, Ruocheng Wang, Mengfei Ran, Hongxu Chen, Junchi Yan Senior Member, IEEE Abstract Simulating quantum many-body systems represents a fundamental challenge where classical machine learning methods are severely bottlenecked by the exponential curse of dimensionality. Variational Quantum Algorithms (VQAs) offer a native paradigm to tackle this by optimizing parameterized unitary evolutions to find the ground states of problem Hamiltonians. However, the efficacy of these VQA is deeply hindered by the challenge of balancing the preservation of critical physical symmetries with the strict constraints of hardware implementability. In this work, we address this dilemma by proposing a hardware-efficient, symmetry-preserving ansatz fortified with complete theoretical guarantees for fermionic systems, termed the Hamming Weight Preserving (HWP) ansatz. We establish the necessary and sufficient conditions for 2-local HWP operators to achieve subspace universality, formally debunking the prevailing assumption that truncation-free simulation requires complex high-order interactions. Empirical validations corroborate our theoretical guarantees, showcasing the exact approximation of arbitrary unitary matrices within the HWP subspace. Crucially, we demonstrate the exceptional versatility of the proposed approach by deploying the exact same ansatz across distinct fermionic models, including diverse molecular electronic structures and the Fermi-Hubbard model. Our proposed HWP ansatz consistently suppresses ground-state energy errors below 1 × 10 − 10 1\times 10^{-10} Ha, achieving a level of precision that surpasses the stringent threshold of chemical accuracy by multiple orders of magnitude. This work establishes a complete, theoretically fortified 2-local framework for symmetry-preserving computation, offering a highly universal and hardware-efficient building block for advancing quantum machine learning and fermionic many-body simulations. I Introduction Simulating quantum many-body systems is the long-standing fundamental challenge across condensed matter physics [ 38 ] , quantum chemistry [ 25 , 13 ] and materials science [ 34 ] , as classical methods are bottlenecked by the curse of dimensionality [ 20 , 33 ] . Given the intrinsic quantum mechanical nature, quantum computers offer a native and mathematically robust paradigm for tackling these simulations [ 20 ] . Within this domain, Quantum Machine Learning (QML) [ 7 , 14 ] , particularly the Variational Quantum Eigensolver (VQE) [ 43 ] , has emerged as the premier algorithm, demonstrating significant potential for achieving practical utility on both near-term and Fault-tolerant devices [ 30 ] . Operating as an ab initio approach [ 25 ] , VQE initializes from a trivial reference state and variationally optimizes a parameterized trial wavefunction (ansatz) to approximate the ground state, thereby extracting critical properties of the many-body system. Therefore, the efficacy of simulating many-body system with VQE is deeply constrained by the design of the variational ansatz [ 14 ] . There have been progressive improvements in variational ansatz design, including hardware-efficient [ 30 ] and Hamiltonian variational types [ 47 , 52 ] . Hardware-Efficient Ansätze (HEA) accommodate near-term devices but blindly explore the unconstrained Hilbert space. They suffer from degraded accuracy and severe barren plateaus since they ignore physical symmetries. Conversely, Hamiltonian-driven frameworks like Unitary Coupled Cluster (UCC) [ 2 , 5 ] inherently preserve symmetries but necessitate high-order, non-local interactions. Compiling these complex terms triggers an explosive growth in circuit depth, rendering them intractable even for anticipated fault-tolerant quantum computing architectures. To navigate the trade-off between hardware efficiency and physical meaning, a more robust approach is to leverage fundamental physical symmetries, which mathematically constrain the state evolution to dynamically isolated subspaces [ 22 ] . For fermionic systems, critical symmetries, such as particle-number conservation, naturally map to the preservation of the quantum-state Hamming weight under standard qubit encodings [ 29 , 8 ] . Existing symmetry-preserving circuit designs predominantly follow a bottom-up strategy, where heuristic operators are derived directly from the symmetry-preserving system Hamiltonian [ 60 , 28 , 4 , 55 , 19 , 52 , 22 ] . However, this fragmented approach lacks any rigorous theoretical guarantee regarding expressivity [ 66 , 36 ] . It remains unknown whether circuits composed of these ad-hoc operators can universally explore the subspace to reach the target ground state. Furthermore, it is conventionally believed that achieving truncation-free simulation of strongly correlated many-body systems strictly necessitates non-local, high-order interaction terms [ 47 , 2 ] . However, compiling such complex operators imposes prohibitive circuit depth overheads, constituting a major bottleneck for the practical deployment of VQEs [ 56 ] . Therefore, discovering a hardware-efficient, symmetry-preserving ansatz fortified with strict theoretical guarantees remains a central and highly practical challenge in QML. In this work, we address this critical gap by presenting a novel top-down framework for constructing mathematically interpretable Hamming Weight Preserving (HWP) ansätze. Diverging from prior bottom-up methods that derive scattered gates from specific Hamiltonians, our paradigm systematically delineates the complete mathematical structure of the HWP space. From this universal foundation, we rigorously deduce the necessary and sufficient conditions required for 2-local operators to achieve complete controllability over the constrained subspace. This top-down derivation renders our ansatz entirely independent of any specific system Hamiltonian, effectively providing a universal, truncation-free building block capable of simulating diverse fermionic many-body systems without resorting to complex high-order terms. Specifically, we first derive the necessary and sufficient conditions for 2-local HWP operators to achieve universality in the HWP subspace by analyzing the Lie Algebra dimension, ensuring maximum expressivity. We then establish the trainability of the ansatz by deriving exact gradient variance expressions [ 39 , 59 ] . This theoretical analysis rigorously quantifies the optimization landscape, demonstrating that the gradient vanishing is strictly bounded by the physical subspace dimension rather than the full unconstrained Hilbert space. The theoretical findings are validated by successfully approximating arbitrary unitary matrices within the HWP subspace [ 42 , 15 , 31 ] , demonstrating the ansatz’s ability to efficiently capture target quantum states or transformations. To further showcase the practical utility of the proposed ansatz, we evaluate on representative Fermionic systems [ 30 , 24 , 27 ] , including solving electronic structures for molecules and the Fermi-Hubbard model. The proposed ansatz achieves remarkable accuracy on both models and suppresses the Hamiltonian variational ansatz in terms of both efficiency and accuracy. II Preliminary II-A Hamming Distance and Hamming Weight Considering two binary vectors \mathbf{a} and \mathbf{b} with \mathbf{a},\mathbf{b}\in\{0,1\}^{N} , where N is the dimension of the vectors. The Hamming distance of these two vectors is as follows: Definition 1 . Hamming distance: The Hamming distance \mathcal{D} of two binary vectors \mathbf{a} and \mathbf{b} is: \displaystyle\mathbf{c}=\mathbf{a}\oplus\mathbf{b}, (1) \displaystyle\mathcal{D}(\mathbf{a},\mathbf{b})=\sum_{i=1}^{N}\mathbf{c}_{i}, where \oplus stands for exclusive OR operator. With the definition of Hamming distance, we can further define the Hamming weight of a given binary vector \mathbf{a} . Definition 2 . Hamming weight: Let \mathbf{0}=0^{N} . The Hamming weight HW of binary vector \mathbf{a} is: HW(\mathbf{a})=\mathcal{D}(\mathbf{a},\mathbf{0}), (2) II-B Lie Algebra and Controllability of Quantum Systems We first introduce a well-defined mathematical tool to derive the theoretical findings, which is Dynamical Lie Algebra (DLA) [ 66 , 18 ] . Lie algebraic techniques have been widely used in discussing the controllability of quantum systems due to the unitary transformation nature of the quantum circuits. For a more intuitive understanding, the Dynamical Lie Group \mathbb{G} comprises unitary matrices, and the corresponding DLA \mathfrak{g} comprises anti-Hermitian matrices. For a L

Read the complete text