Molecular simulations are essential in computational chemistry, with classical and quantum methods offering complementary approaches to studying molecular structures and interactions. This thesis explores two distinct projects: classical geometry optimization (Chapter 2) and quantum computing for electronic structure simulations (Chapters 3 and 4). Chapter 2 investigates geometry optimization using various open-source optimizers interfaced with the QUICK program. By analyzing structures from the Baker test set, we compare internal and Cartesian coordinates, evaluate quasi-Newton strategies, and assess a machine learning-based Gaussian Process Regression (GPR) optimizer. Among the tested methods, ASE/Berny and ASE/Sella achieve the fastest convergence, making them suitable for large-scale applications. Chapter 3 presents quantum-centric simulations of noncovalent interactions using sample-based quantum diagonalization (SQD). We compute the potential energy surfaces of water and methane dimers on quantum hardware, achieving deviations within 1 kcal/mol from high-accuracy classical methods. A 54-qubit simulation explores the current limitations of quantum methods in modeling hydrophobic interactions. Chapter 4 extends quantum-centric simulations to larger systems by integrating SQD with density matrix embedding theory (DMET). This hybrid approach efficiently reduces quantum resource requirements while maintaining accuracy. Simulations of an 18-atom hydrogen ring and cyclohexane conformers on the ibm\_cleveland quantum device demonstrate the feasibility of quantum embedding techniques for extended molecular systems. By advancing classical optimization strategies and demonstrating scalable quantum simulations, this thesis contributes to the development of computational tools for accurate and efficient molecular modeling.
Read
