Density functional theory (DFT) has become the most widely used first-principles computational method to simulate different atomic, molecular, and solid phase systems based on electron density assumptions. The complexity of describing a many-body system has been significantly reduced in DFT. However, it also brings in potential error when dealing with a system that involves the interactions between metallic and non-metallic species. DFT tends to overly-delocalize the electrons in metallic species and sometimes results in the overestimation of reaction energy, metallic properties in insulators, and predicts relative surface stabilities incorrectly in some instances.
There are two approaches to overcoming the failure of DFT using standard exchange-correlation functionals: One can either use a higher level of theory (and thus incur a greater computational cost) or one can apply an efficient correction scheme. However, inaccurate corrections and improper calculation models can also lead to more errors. In the beginning of this dissertation, we introduce the correction methods we developed to accurately model the structure and electron density in material surfaces; then we apply the new methods in surface reactivity studies under experimental conditions to rationalize and solve real life problems.
We first investigate the post-DFT correction method in predicting the chemisorption energy (Echem) of a NO molecule on transition metal surfaces. We show that DFT systematically enhances back-donation in NO/metal chemistorption from the metal d-band to NO 2π* orbital, and relate the back-donation charge transfer to the promotion of an electron from the 5σ orbital to the 2π* orbital in the gas-phase NO G2Σ-←X2Π excitation. We establish linear relationships between Echem and ΔEG←X and formulate an Echem correction scheme to the (111) surfaces of Pt, Pd, Rh and Ir.
As a precursor to further optimization of DFT corrections on transition metal oxide surfaces, we systematically compare the alumina (α-Al2O3) and hematite (α-Fe2O3) (0001) surfaces to study how the atomic positions treatment during geometry optimizations would affect the electronic structure and modeled reactivity, since they are often reported to have a minimal effect. Our results suggest that both can vary significantly in quantitative and qualitative ways between partially constrained or fully relaxed slab models.
We continue to use the α-Fe2O3 (0001) surfaces to optimize the Hubbard U method implemented in DFT that determines the Coulomb repulsion correction (Ud) to localize Fe d-electrons. It successfully restores the insulating properties of bulk hematite, but underestimates the stability of the oxygen-terminated surface. It is mainly due to the fact that all the chemically distinct surface Fe atoms were treated the same way. Here we develop a linear response technique to derive specific Ud values for all Fe atoms in several slab geometries. We also find that in a strongly correlated system, the O p-orbitals also need the Hubbard correction (Up) to accurately predict the structural and electronic properties of bulk hematite. Our results show that the site-specific Ud, combined with Up as Ud+p, is crucial in obtaining theoretical results for surface stability that are congruent with the experimental literature results of α-Fe2O3 (0001) surface structure.
Besides methodology development, we continue to apply our specific Ud+p method in the engineered application of the Chemical Looping Combustion (CLC) process in which transition metal oxides play the role of oxidizing fuel molecules for full CO2 capture. Current molecular dynamic studies use partially constrained surface models to simulate the CH4 reaction on hematite surfaces without the detailed comparison of the early stage adsorption products. Here we use hematite (α-Fe2O3) and magnetite (Fe3O4) surfaces as analogous to systematically study the early adsorption products of CH4. Our results show that the reaction favors the homolytic pathway on O-terminated surface, and that as a reduced form of hematite, the magnetite surface also shows excellent reactivity on CH4 dissociation.
Knowing how to simulate DFT surface model properly we continue to enrich our theoretical methods for more complicated systems under aqueous conditions. We focus on various structures of the lithium-ion battery material, LiCoO2 (LCO) (001) surface, involving hydroxyl groups. We assess the relative stabilities of different surface configurations using a thermodynamic framework, and a second approach using a surface-solvent ion exchange model. We find that for both models the –CoO–H1/2 surface is the most stable structure near the O-rich limit, which corresponds to ambient conditions. We also found that this surface has nonequivalent surface geometry with the stoichiometric –CoO–Li1/2 surface, leading to distinct band structures and surface charge distributions. We go on to probe how those differences affect the surface reactivity in phosphate anion adsorption.
All of the work presented in this dissertation reveals the importance of accurately modeled material structures in theoretical studies to achieve correct physical properties and surface reactivity predictions. We hope our DFT correction schemes can continue to contribute to future surface studies and experimental measurements, and to enlighten new ideas in future DFT methodology improvements.