Degree

Bachelor of Science (Computer Science)

Department

Department of Computer Science

School

School of Mathematics and Computer Science (SMCS)

Advisor

Dr. Jibran Rashid, Assistant Professor, Department of Computer Science

Keywords

Quantum Computing, Quantum Chemistry, Variational Algorithms

Abstract

The Variational Quantum Eigensolver (VQE) is a hybrid quantum-classical algorithm tailored to the Noisy Intermediate-Scale Quantum (NISQ) era, with its principal application in quantum chemistry: estimating molecular ground-state energies, which underpins the understanding of chemical properties. While exact classical methods such as Full Configuration Interaction scale exponentially with system size, VQE leverages variational principles and parameterized quantum circuits to approximate these quantities with potential advantages on suitable hardware. This project began as a comprehensive benchmarking study of the full VQE pipeline across basis sets, fermion-to qubit mappings, ansatzes, optimizers, and backends. Through initial experimentation and consultation with researchers in the field, it became clear that optimal pipeline configurations are molecule-specific and do not generalize, making broad benchmarking a less productive direction than anticipated. The project therefore pivoted to a more focused contribution: investigating the practical resource efficiency of ADAPT-VQE on NISQ hardware by combining multiple methods and advancements made on ADAPT-VQE, to demonstrate the practical viability of adaptive ansatz construction on real quantum hardware. Specifically focusing on integrating the Coupled Exchange Operator (CEO) pool with the Overlap-ADAPT-VQE selection criterion and the TETRIS-ADAPT-VQE framework for shallower circuit construction.

Tools and Technologies Used

Python, Qiskit, PennyLane, Jax, OpenMP

Methodology

The project was conducted in two stages reflecting its change in direction. In the first stage, a broad benchmarking pipeline was implemented and evaluated on the hydrogen molecule, sweeping combinations of basis sets, fermion-to-qubit mappings, ansatzes, and optimizers to identify high-performing configurations. This stage established the core infrastructure and produced baseline results, but also revealed a fundamental limitation: configurations that performed well on H2 did not reliably transfer to other molecules, and no consistent general-purpose configuration emerged. Following discussions with researchers in the field and the project supervisor, the scope was narrowed to a focused investigation of hardware resource efficiency within the ADAPT-VQE framework. The second stage follows a three-configuration experimental design built around a shared Overlap-Vanishing Paired Compact Excitation Operator (OVP-CEO) pool, ensuring that performance differences are attributable solely to the selection criterion and ansatz construction strategy. The baseline configuration runs CEO-ADAPT-VQE with standard gradient-based selection. The second configuration replaces the selection criterion with the overlap-based approach of Feniou et al., running an initial phase that builds the ansatz by maximizing overlap with a pre-computed reference state before switching to energy gradient selection for the remaining iterations. The third configuration extends this by additionally applying the TETRIS framework, selecting multiple operators with disjoint qubit supports per iteration to reduce circuit depth without increasing CNOT count. All three configurations are evaluated on CH4 and H2O at stretched bond geometries using the LinAlg exact statevector simulator, with performance measured in terms of operators and CNOT gates required to reach chemical accuracy of 1.6 mHa from the FCI reference energy.

Document Type

Restricted Access

Submission Type

BSCS Final Year Project

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Download

Share

COinS