Energy storage systems, in particular, those based on electrochemical reactions, are becoming more and more crucial to power an environmentally friendly world with renewable energies. As such, sustainable energy technologies with high efficiency and fuel flexibility are needed to curb global warming and limit environmental pollution. Solid oxide fuel cells (SOFCs) stand out due to their high efficiency and fuel flexibility. Particularly, SOFCs can operate with several fuels, including H₂, natural gas, propane, etc. Despite the great promise, the SOFC technology has not been commercialized mainly due to the high operating temperatures. Lowering the temperature below 800 °C brings another limitation that the oxygen reduction reactions (ORR) taking place at the cathode side of the...[ Read more ]

Energy storage systems, in particular, those based on electrochemical reactions, are becoming more and more crucial to power an environmentally friendly world with renewable energies. As such, sustainable energy technologies with high efficiency and fuel flexibility are needed to curb global warming and limit environmental pollution. Solid oxide fuel cells (SOFCs) stand out due to their high efficiency and fuel flexibility. Particularly, SOFCs can operate with several fuels, including H₂, natural gas, propane, etc. Despite the great promise, the SOFC technology has not been commercialized mainly due to the high operating temperatures. Lowering the temperature below 800 °C brings another limitation that the oxygen reduction reactions (ORR) taking place at the cathode side of the SOFCs are typically sluggish. This shortcoming may be overcome by applying the mixed ionic-electronic conductors (MIEC), which are capable of extending the active reaction zone beyond the triple-phase boundary to the entire two-phase boundary. BaFeO_3-δ (BFO), as one MIEC, has recently attracted much research attention. However, the conductivity of these materials is typically low.

In this thesis, we first showed by density functional theory (DFT) calculations that substituting Fe in Ba_1-xLa_xFeO_3-δ with P lowers both the O vacancy formation and vacancy migration energies and promotes the creation of PO4 groups. These factors contribute to improving diffusional properties and oxygen reduction reaction performance. We further successfully introduce P in the Fe site of Ba_0.95La_0.05FeO_3-δ (BLF) to make a novel cathode material, Ba_0.95La_0.05Fe_0.95P_0.05O_3-δ (BLFP). Consistently, we observe that BLFP has higher electrical conductivity and better electrocatalytic activity compared to BLF. Another factor that hinders the BLF activity is the surface Ba enrichment along with the Fe depletion, causing a drastic reduction of the electrocatalytic activity. However, a deep understanding of the segregation mechanism of Ba toward the surface is still absent. Using slab models, we further studied the thermodynamic surface stability of BaFeO₃ using density functional theory calculations and modeled its behavior at different conditions of temperature and pressure. We found that the (001) BaO-terminated surface is the most stable one in a wide range of temperatures (𝑇 ≤ 1000 K) at ambient pressure, justifying that the experimentally observed Ba surface enrichment may be caused by the dominating role of (001) BaO-termination. Further calculations suggest that the surface composition can be well controlled by tuning the lattice strain, and a tensile strain can stabilize the (110) BaFeO-termination, thus lowering the surface Ba/Fe ration. It was noticed that the surface stability correlates with the Ba-O bond length, implying a longer Ba-O bond leads to better stability of the (110) BaFeO-termination, thus affecting the surface composition. This hypothesis was verified by substituting the Fe with Zr to elongate the Ba-O bond. The simulation results of Zr-substituted BaFeO₃ confirmed that the stability of the (110) BaFeO surface had been strengthened. As such, we concluded that the lattice strain, or equivalently the Ba-O bond length, plays a key role in controlling the surface thermodynamic phase diagram of BFO and BFZ, thus the surface compositions.

In addition to the SOFCs, another example of the commonly used electrochemical energy storage systems, alkaline ion battery, e.g., Li-ion battery (LIB) or Na-ion battery (NIB), has gained much development in the past decades. Despite the great potential, conventional LIBs and NIBs still face several challenges, e.g., the low energy density, and the flammability of the liquid electrolyte. These two challenges may be addressed by replacing the liquid electrolytes with solid-state ones. In the fourth chapter of this thesis, we discussed the design strategies of a new solid-state electrolyte using the first-principles calculations. By substituting Sn with Ge in Na₁₁Sn₂PS₁₂, we were able to improve the ionic conductivity significantly. This work may shed light on designing next-generation solid-state batteries.

To characterize physicochemical processes taking place in these systems, electrochemical impedance spectroscopy (EIS) is one of the most commonly used approaches as it can cover a much broader timescale of electrochemical processes of the systems. However, interpreting the EIS spectra is not easy. The equivalent circuit model (ECM) and physical model are two primarily applied frameworks. Nonetheless, they both face significant challenges. In specific, different ECMs may lead to equivalent fitting performance, adding the difficulty in further explanation. Physical models are developed for specific problems. Distribution of relaxation time (DRT) is one generalized approach that gets rid of the equivalent circuits, and it also directly assesses the characteristic time for different processes. However, deconvolving the EIS to obtain the DRT is always an ill-posed problem. In this thesis, we propose three different approaches to obtain the DRT given the EIS spectra. The first one is based on the Gaussian process (GP), where the DRT is assumed to be a GP. By maximizing the likelihood of the experimental data, the DRT can be inferred. This allows us to get the credible intervals for the measured impedance and predict the impedance values at unmeasured frequency points. The second one uses the deep neural network (DNN) to automatically optimize the DRT directly. At last, we reformulated the Hilbert transformed DRT under the Bayesian framework, which allows us to score the compliance of the impedance data with Kramers-Kronig relations.