Gradient-Based Excitation Filter for Molecular Ground-State Simulation
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911040979075072 |
|---|---|
| author | He, Runhong Chai, Qiaozhen Hong, Xin Guan, Ji Cui, Guolong Wang, Shengbin Ying, Shenggang |
| author_facet | He, Runhong Chai, Qiaozhen Hong, Xin Guan, Ji Cui, Guolong Wang, Shengbin Ying, Shenggang |
| contents | Molecular ground-state simulation is one of the most promising fields for demonstrating practical quantum advantage on near-term quantum computers. However, the Variational Quantum Eigensolver (VQE), a leading algorithm for this task, still faces significant challenges due to excessive circuit depth. This paper introduces a method to efficiently simplify the Unitary Coupled-Cluster with Single and Double Excitations (UCCSD) ansatz on classical computers. We propose to estimate the correlation energy contributions of excitations using their gradients at Hartree-Fock state, supported by a theoretical proof. For molecular systems with $K$ orbitals, these gradients can be obtained with complexity only $O(K^8)$, which can be efficiently implemented on classical computers, especially in parallel. By sorting and truncating the excitations based on these gradients, the simplified ansatz can be obtained immediately, avoiding the challenging task of optimizing ansatz structure on a quantum computer. Furthermore, we introduce a strategy to indirectly identify critical excitations through spin-adapted constraints, reducing gradient computations by $60\%$. Numerical experiments on prototype molecular systems (H${_4}$, HF, H${_2}$O, BeH${_2}$ and NH$_3$) demonstrate that our approach achieves up to $46\%$ parameter decrease, $60\%$ circuit depth reduction and $678\times$ runtime speedup compared to the state-of-the-art ADAPT-VQE algorithm, enabling significantly more compact quantum circuits with enhanced near-term feasibility. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_20398 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Gradient-Based Excitation Filter for Molecular Ground-State Simulation He, Runhong Chai, Qiaozhen Hong, Xin Guan, Ji Cui, Guolong Wang, Shengbin Ying, Shenggang Quantum Physics Molecular ground-state simulation is one of the most promising fields for demonstrating practical quantum advantage on near-term quantum computers. However, the Variational Quantum Eigensolver (VQE), a leading algorithm for this task, still faces significant challenges due to excessive circuit depth. This paper introduces a method to efficiently simplify the Unitary Coupled-Cluster with Single and Double Excitations (UCCSD) ansatz on classical computers. We propose to estimate the correlation energy contributions of excitations using their gradients at Hartree-Fock state, supported by a theoretical proof. For molecular systems with $K$ orbitals, these gradients can be obtained with complexity only $O(K^8)$, which can be efficiently implemented on classical computers, especially in parallel. By sorting and truncating the excitations based on these gradients, the simplified ansatz can be obtained immediately, avoiding the challenging task of optimizing ansatz structure on a quantum computer. Furthermore, we introduce a strategy to indirectly identify critical excitations through spin-adapted constraints, reducing gradient computations by $60\%$. Numerical experiments on prototype molecular systems (H${_4}$, HF, H${_2}$O, BeH${_2}$ and NH$_3$) demonstrate that our approach achieves up to $46\%$ parameter decrease, $60\%$ circuit depth reduction and $678\times$ runtime speedup compared to the state-of-the-art ADAPT-VQE algorithm, enabling significantly more compact quantum circuits with enhanced near-term feasibility. |
| title | Gradient-Based Excitation Filter for Molecular Ground-State Simulation |
| topic | Quantum Physics |
| url | https://arxiv.org/abs/2506.20398 |