Improving Quantum Simulation Efficiency Using Hardware-Adaptable Ansatz Based on Quantum Neural Networks Syawal Adrian Syah, Teguh Budi Prayitno, Yanoar Pribadi Sarwono
Physics Department, Faculty of Mathematics and Natural Sciences, Jakarta State University
Quantum Physics Research Center, National Research and Innovation Agency
Abstract
The current limitations of quantum devices remain a major challenge, including the restricted number of qubits, high levels of noise, and decoherence. These constraints demand the development of efficient quantum circuits (ansatz) to ensure that variational algorithms such as the Variational Quantum Eigensolver (VQE) can operate reliably on near-term quantum hardware. This study aims to develop an effective Hardware-Adaptable Ansatz-Quantum Neural Network (HAA-QNN) architecture. The proposed ansatz leverages ancilla qubits to reduce the need for deep circuit layers, while optimizing its structure and parameters to enhance quantum simulation efficiency. An experimental approach is employed, wherein the performance of the HAA-QNN is tested and evaluated against other ansatz, including the Hardware-Efficient Ansatz (HEA) and the qubit-reused QNN (qrQNN). The evaluation is conducted through ideal simulations using Qiskit and Pennylane, based on three key metrics: ground state energy, Root Mean Square Error (RMSE), and expressibility. The simulation results indicate that HAA-QNN achieves the fastest and most stable convergence toward the ground state energy, yielding final values closest to the exact solution obtained via Full Configuration Interaction (FCI). Furthermore, HAA produces significantly lower RMSE compared to both HEA and qrQNN, and demonstrates high expressibility as indicated by a notable reduction in KL divergence. In addition, HAA responds more efficiently to increased model capacity than the other ansatz. Therefore, HAA-QNN is proven to be a superior ansatz architecture for VQE, offering advantages in energy accuracy, convergence efficiency, and representational flexibility.
