Di Grande, Silvia (2025) Towards accurate thermochemical properties: computational methodologies for molecular systems of increasing dimensions. [Tesi di dottorato]

Documento PDF DiGrande_Silvia_37.pdf Download (18MB)

Abstract

The quest for ever-increasing accuracy remains one of the central challenges in quantum chemistry, motivating the development of methods that aim to achieve a one-to-one correspondence between experiments and first-principles models. However, the most accurate approaches are generally feasible only for small molecules, while methods applicable to large systems rely on approximations that limit their predictive power. This thesis tackles this limitation by developing new composite approaches that combine accuracy and scalability, designed to bridge the gap between the size of the molecular systems investigated and the level of reliability of simulated properties. This objective is pursued by developing three closely related research lines. First, the junChS-F12 model is improved, leading to a composite scheme (CS) capable of providing accurate thermochemical and rotational–vibrational spectroscopic properties for medium-size molecules containing elements from the first three rows of the periodic table. This advancement has improved the predictive power of the existing explicitly correlated ChS framework and extended its applicability beyond energetics to equilibrium geometries and harmonic vibrational frequencies. The results have established the junChS-F12 as a robust and consistent CS, forming a solid foundation for subsequent methodological developments. Building on these achievements, in a second step, the Pisa Composite Scheme (PCS) family is introduced a new and modular generation of composite protocols offering higher accuracy than the ChS while retaining similar computational demands. Within this context, the core protocol of the PCS series has been designed for molecular systems containing approximately 5–15 atoms, with the energetic and structural schemes rooted in explicitly correlated CCSD(T) methods, which have served as the workhorses for this molecular size regime. The approach is then been extended to larger systems, up to about 50 atoms, while maintaining the same methodological philosophy. In this case, energetic contributions have been evaluated using the frozen natural orbital (FNO) approximation within the Coupled Cluster (CC) formalism, whereas geometries have been determined through DFT-based schemes. In both methodological variants, the computational level has been progressively tailored to the computational cost of each target property and to the characteristics of the examined molecular system. Finally, with the aim of extending accurate composite methodologies to larger molecular systems at reduced computational cost, post-CCSD(T) correlation effects are accounted for through the FNO approximation. This approach has allowed the computation of high-order correlation contributions for molecular systems beyond the reach of canonical CC methods, significantly reducing computational requirements while maintaining accuracy. This methodology has thus opened the way toward spectroscopic accuracy for molecular sizes that were previously unattainable due to the prohibitive cost of higher-order CC excitations. Overall, this research demonstrates that the integration of state-of-the-art wave function-based techniques within carefully designed composite frameworks enables the development of predictive quantum chemical models that are both accurate and computationally scalable, effectively reducing the gap between benchmark-level methods and molecular systems spanning different size regimes, with applications ranging from astrochemical to biochemical and technological contexts.

Downloads

Actions (login required)

Modifica documento