Solid-state synthesis prediction is a key accelerator for the rapid design of advanced inorganic materials. However, determining synthesis variables such as the choice of precursor materials is challenging for inorganic materials because the sequence of reactions during heating is not well understood. To achieve predictive synthesis for the desired material, one potential approach is to learn synthesis design patterns from a large volume of experimental synthesis procedures. Nevertheless, a comprehensive, large-scale database of structured synthesis procedures for inorganic materials does not exist. Provided the ability of converting unstructured text to structured information, the decades of solid-state chemistry literature constitutes a treasure trove of synthesis data. Therefore, this study aims at: (1) developing natural language processing (NLP) algorithms to text mine a large-scale inorganic synthesis dataset from materials science literature, and (2) developing machine learning algorithms for precursor selection in solid-state synthesis based on the text-mined dataset.

Although many general-purpose NLP methods exist, text mining for inorganic synthesis requires dedicated development of models for information retrieval (Chapter 2). During the development of a text-mining pipeline, one major problem is the difficulty of identifying which materials from a synthesis paragraph are precursors or are target materials. In this study, we developed a two-step Chemical Named Entity Recognition (CNER) model to identify precursors and targets, based on information from the context around material entities. By integrating our information retrieval model for precursors and targets, and also the ones for other synthesis variables, we established a fully automated text-mining pipeline that extracts the structured data of synthesis procedures from the literature. Starting from 4,973,165 materials science papers, we applied our text-mining pipeline and successfully extracted 33,343 solid-state synthesis procedures. The quality of the text-mined synthesis dataset is validated by the high accuracy of 93% at the chemistry level, where each extracted reaction has the target and precursor materials consistent with the original literature report. This dataset for inorganic solid-state synthesis is currently the largest of its kind and paves the way toward the development of data-driven approaches for rational synthesis design.

Using the extracted data, we conducted a meta-analysis to study the similarities and differences between precursors in the context of solid-state synthesis (Chapter 3). To quantify precursor similarity, we built a substitution model to calculate the viability of substituting one precursor with another while retaining the target. From a hierarchical clustering of the precursors, we demonstrate that the "chemical similarity" of precursors can be extracted from text data, without the need to include any explicit domain knowledge. Quantifying the similarity of precursors offers a reference for suggesting candidate reactants when researchers alter existing recipes by replacing precursors. The capability of creating alternative recipes constitutes an important step toward developing a predictive synthesis model.

While the selection of alternative precursors is enabled by the similarity of precursors, it is limited to existing materials. To learn which precursors to recommend for the synthesis of a novel target material, we further developed a representation learning model to evaluate the similarity of targets (Chapter 4). The data-driven approach learns "chemical similarity" of target materials and refers the synthesis of a new target to precedent synthesis procedures of similar target materials, mimicking human synthesis design. When proposing five precursor sets for each of 2,654 unseen test target materials, our recommendation strategy achieves a success rate of at least 82%. Our approach captures decades of heuristic synthesis data in a mathematical form, making it accessible for use in recommendation engines and autonomous laboratories.

Overall, this study contributes a valuable large-scale synthesis dataset and interpretable precursor selection algorithms to the materials science community, representing a step forward in the prediction of solid-state synthesis.

Disordered rocksalt Li-excess (DRX) compounds form a potentially new class of cathode materials for Li-ion batteries, as they contain only abundant metals and do not require Co or Ni. While DRX compounds are promising as battery materials, their properties and performance are not as optimized as those of conventional layered cathode materials. To compete with conventional cathodes, developing a more detailed understanding of the structure-property relation in DRX compounds as it relates to their electrochemical applications is necessary. Specifically, the Li transport properties in DRX materials are not well understood, despite their significant impact on the rate performance of the battery. This dissertation employed first-principles calculations, the cluster expansion technique, and the kinetic Monte Carlo method to investigate the Li transport properties of DRX materials. Chapter 1 provides an introduction to DRX, including a detailed discussion of the possible Li+ diffusion mechanisms and percolation in these materials. Chapter 2 investigates the impact of fluorine and Li-excess on Li migration barriers in rocksalt structures. Fluorine substitution in the oxygen sublattice reduces oxygen redox and enhances energy density and capacity retention. However, its impact on Li transport remains unclear. Using first-principles calculations, this chapter explores the effects of both F substitution and accompanying Li-excess on Li migration barriers. The results show that F has a small negative effect on Li migration barriers while Li-excess decreases Li+ migration barriers. Because fluorination enables more Li-excess, these results do not predict any detrimental impact on Li transport. Chapter 3 delves into the essential aspects of the delithiation process and cation short-range ordering (SRO) in DRX compounds. This study uses first-principles calculations and the cluster expansion approach to model the disorder in DRX Li2-xVO3 with 0 ≤ x ≤ 1. The chapter discusses the SRO of Li in tetrahedral and octahedral sites, and the order in which Li delithiates and V oxidizes with respect to local environments. The results indicate that the number of nearest neighbors V determines the order of octahedral Li delithiation, and that V are oxidized in a manner that minimizes the electrostatic interactions among V. This research provides valuable insights into how to tailor the performance of V-based DRX cathode materials by controlling SRO features that may reduce energy density. Additionally, findings of this chapter shed light on the relationship between SRO and Li transport properties. Chapter 4 investigates the kinetics of Li+ transport in disordered rocksalt structures, focusing on DRX Li2-xVO3. Using a combination of first-principles calculations, cluster expansion approach, and kinetic Monte Carlo simulations, the chapter provides insights into the factors affecting Li+ transport properties in DRX materials. The results show that the relative stability of tetrahedral Li and octahedral Li occupancy plays a crucial role in determining the number of active sites in the percolation network and the resulting Li+ transport properties. Furthermore, the chapter demonstrates that the wide site-energy distribution can cause correlated motion in Li2-xVO3, which can hinder Li+ transport. Although this study focuses on Li2-xVO3 as a model system, the insights gained apply to all DRX materials, given their inherently broad site-energy distributions. In the concluding chapter, Chapter 5, the key findings of this dissertation are summarized.

As the use of lithium-ion batteries in consumer electronics and electric vehicles has grown, safety issues such as those arising from leakage and flammability of the organic liquid electrolyte have garnered increased attention. By replacing the organic liquid electrolyte with an inorganic solid electrolyte, however, these concerns can be circumvented, thereby improving the safety of the battery system. In addition, solid-state batteries are also expected to possess higher energy density than their conventional counterparts. On the anode side, the high modulus of the solid electrolyte is expected to constrain the growth of Li dendrites, which might enable the use of a Li metal anode. Furthermore, because some solid electrolytes possess a wide electrochemical stability window, high-voltage cathodes (> 4.5 V, vs Li+/Li) may be used to further improve the total energy density.

The superionic conductor is one of the key parts of solid-state batteries. Over the past 20 years, accelerated development of Li superionic conductors has occurred. The ionic conductivities of some of these superionic conductors approach or even surpass those of liquid electrolytes. However, most reported superionic conductors have obvious drawbacks. New superionic conductors that meet all the requirements of solid-state batteries are needed. There are two strategies for exploring new potential superionic conductors: (1) modification the chemical composition based on the crystal structure of known fast Li conductors to further improve the properties or (2) searching for Li conductors with new crystal structures based on the structural features that favor fast Li+ migration. In this dissertation, new superionic conductors are designed and explored by implementing both of these strategies.

Specifically, a strategy is developed to increase the ionic conductivity of sulfide Li-ion conductors through composition modification. Inspired by the wide use of halogens in superionic conductors, we propose that the conductivity could be further improved by substituting halogens with suitable pseudo-halogens. The Li argyrodite system was used to demonstrate the feasibility of this strategy. BH4-substituted Li argyrodite was successfully synthesized and shown to have a room-temperature ionic conductivity of 4.8 mS/cm which is 5 times higher than that of halogen-substituted Li argyrodites. We further discuss the mechanism underlying the enhanced ionic conductivity and find that the faster Li diffusion originates from the weak interaction between Li and BH4. The results provide design strategies for new superionic conductors with pseudo-halogen substitution.

We also present a structural feature that benefits the Li-ion migration in oxide Li-ion conductors. Based on a statistical analysis of the materials in the inorganic materials database, we discovered that the corner-sharing connectivity of the oxide crystal structure framework is more likely to have a distorted lithium environment with higher site energy. Materials with a corner-sharing framework are also usually less compact, which reduces the repulsion from non-lithium cations. Both features lead to a decreased migration barrier and accelerate the Li diffusion. A high-throughput search was performed based on this structural feature, and 10 new oxide Li-ion conductors were predicted. One of them, LiGa(SeO3)2, was successfully synthesized and was shown to have a bulk conductivity of 0.11 mS/cm, in agreement with theory predictions. These findings provide fundamental insights into the physical attributes that govern fast lithium conduction and help project new directions towards the discovery of superionic conductors for all-solid-state batteries.

Lithium-ion batteries have achieved tremendous success in energy storage with a broad range of application. The Li-ion industry is growing to 1TWh of production per year for which approximately one million tons of combined cobalt or nickel will be required. The increasing demand has imposed a large strain on nature resources. Compared to traditional commercialized layered Li(Ni,Mn,Co)O2 (NMC) cathodes, which require expensive cobalt and nickel, the recent development of Li-excess cation-disordered rocksalt (DRX) cathodes widely broaden the possible chemical space from which to draw Li-ion cathodes, including some inexpensive and earth-abundant TMs such as Mn, Ti, Fe, V and Cr. Unlike the well-defined 2-D percolation pathway in layered NMC cathodes, the Li percolation in DRX structures relies on the 3-D quasi-randomly distributed 0-TM channels (no transition metal). Although long-range disordered, the DRX structure still contains short-range cation order (SRO), which has significant impact on Li percolation and electrochemical performance. Therefore, extensive research efforts are required to achieve a better understanding of the correlation between cation order and overall electrochemical performance. In this thesis, three strategies to modify local ordering in DRX cathodes, including metal doping on cation sublattices, synthesis condition control and cation partially order, were carefully evaluated by combining electrochemical tests, advanced characterizations and computational investigations.The first part of the thesis will focus on the cation doping strategy to control the SRO and extractable Li in fluorinated DRX cathodes. Due to the ‘locking effect’ caused by strong bonding between Li and F, not all the Li can be extracted during charging, which compromises the electrochemical performance. A screening based on the bonding strength and ionic radius was performed to identify a suitable dopant to free the locked Li. Mg was chosen because of its large bonding energy with F and similar ionic radius with Li. A detailed theoretical investigation of Li–F SRO using DFT calculations and cluster expansion (CE) Monte Carlo (MC) simulations confirmed that Mg doping results in a greater fraction of extractable Li ions by reducing the frequency of the Li6F configuration. We verified our hypothesis by comparing the electrochemical performance of Li1.25Mn0.45Ti0.3O1.8F0.2 and Li1.25Mg0.1Mn0.45Nb0.2O1.8F0.2. Although both compounds have the same Li-excess and Mn content, the Mg-doped compound delivers higher capacity. Finally, we extended the discussion to a related group of compositions: Li1.333Mn0.667O1.333F0.667 (LMF), Li1.233Mg0.1Mn0.667O1.333F0.667 (ls-LMF), Li1.333Mg0.1Mn0.567O1.333F0.667 (ms-LMF) and Li1.28Mg0.11Mn0.61O1.333F0.667 (ls-LMF) to demonstrate that increased accessible Li, enabled by Mg doping, can also be traded for an improvement of cycle life. The results showed that Mg-doping should be considered in fluorinated DRX cathodes to optimize the performance. The second part of the thesis reports on the solid-state synthesis mechanism of DRX compounds. The synthesis mechanism of a typical DRX oxyfluoride, Li1.2Mn0.55Ti0.25O1.85F0.15 (LMTF), are characterized to investigate the formation of long-range order and short-range order. Experimentally, the formation of long-range disorder followed by short-range order during the synthesis was clearly demonstrated through in-situ transmission electron microscopy, ex-situ X-ray diffraction and pair distribution function analysis. This phenomenon is further explored by performing ab initio calculations for the formation energy of DRX with random structure and with SRO to estimate the reaction kinetics. Based on both experiment and computational results, the SRO feature can be eliminated by simply shortening sintering time and quenching from high temperature. With this strategy, the SRO-free LMTF (35 min) sample exhibited a capacity of >310 mAh g−1 and a specific energy close to 1000 Wh kg−1 and enabled discharge up to 2 A g−1, which is much better than SRO-rich LMTF (4h). The third part of the thesis identifies the cation to anion ratio in synthesis as a key parameter for tuning the structure continuously from a well-ordered spinel, through a partially ordered spinel, to a rocksalt. A series of well selected partially-(dis)ordered spinel cathodes with different degrees of cation over-stoichiometry, Li1.4+xMn1.6O3.7F0.3 (x = 0.07, 0.28, and 0.6), was synthesized via a mechanochemical method. The degree of cation disorder continuously varies with cation to anion ratio, which modifies the voltage profile, rate capability, and charge-compensation mechanism in a rational and predictable way. The results indicate that spinel-type order is most beneficial for achieving high-rate performance as long as the cooperative Li migration from 8a to 16c sites is suppressed. We also found that more rocksalt-like disorder facilitates O redox, which can increase capacity. These findings reveal an important tuning handle for achieving high energy and power in the vast space of partially ordered cathode materials.

Solid-state batteries (SSBs) have received considerable attention due to their reduced flammability and the ability to use high-energy-density metal anodes compared with those of liquid-electrolyte battery systems. Among known inorganic solid electrolytes (SEs), thiophosphates offer the advantage of a combination of high ionic conductivity, low cost and ease of processing, making them strong candidates for large-scale applications. However, thiophosphate electrolytes suffer from poor stability, leading to unstable interfaces upon contact with electrodes. The (electro-)chemical stability of various interfaces in SSBs has become the bottleneck for stable cycling of SSBs.

Thiophosphates are subjected to Li extraction under high voltages. In this thesis, we present a methodology that combines computational and experimental tools to determine the voltage windows of the electrolytes (both the thermodynamic-driven lower bound and kinetic-stabilized upper bound). We demonstrate that sufficient contact between the solid electrolyte and electron source(s) is particularly important in identifying the electrolyte stability window using electrochemical methods. The intrinsic instability of solid electrolyte is related to the capacity loss in solid state batteries, in particular during the first cycle. In this thesis, we also develop a methodology based on in situ synchrotron X-ray diffraction techniques to analyze the capacity fade due to electrolyte decomposition in a full-cell SSB.

A thermodynamically stable cathode–electrolyte pair is rare in SSBs. A chemical reaction between the cathode and SE can occur during battery cycling. In this thesis, we employ relatively simple and rapid XRD and SDT experiments to describe the interfacial reactions between electrolytes and electrodes. With the aid of high-throughput density functional theory calculations, we screen the compatibility between SEs and cathodes in a large chemical space and understand the thermodynamics and kinetics behind the reactions between incompatible electrolyte/electrode pairs.

To suppress the reaction between the cathode and thiophosphate SEs, various cathode coating materials have been developed to serve as a buffer layer and prevent direct contact between the cathode and SE. In this thesis, we directly observe the severity of the interfacial reactions between a thiophosphate electrolyte (Li2S-P2S5) and oxide cathode (NCM) during normal battery operations using different microscopy techniques. A new borate (Li3B11O18) is developed and exhibits excellent oxidation stability and chemical stability, leading to substantially improved performance over cells with Li2ZrO3–coated or uncoated cathodes.

The use of Li/Na metal anodes is essential for SSBs to achieve a high energy density. In this thesis, we demonstrate the degradation of the solid electrolyte results in the formation of mixed electronic and ionic conductive decomposition, leading to significantly increased overall cell impedance and poor rate performance. With the aid of first principles computations, we identify passivating products to stabilize the highly reactive interface between a thiophosphate electrolyte (Na3SbS4) and Na metal. A passivating interface is constructed through introduction of elements and/or compounds that beneficially reacts with Na metal and forms the passivating products. A general strategy of interface design to stability the electrolyte/metal interface is proposed.

The emerging electrical vehicle market and global decarbonization trend demand batteries with higher energy density, better safety, and longer cycle life. An all-solid-state battery has the potential to meet such requirements and is widely considered as the next-generation battery technology to replace the current Li-ion batteries. The remarkable success in the discovery of ceramic alkali superionic conductors has provided a wide selection of solid electrolytes, allowing the fabrication and testing of all-solid-state battery cells. However, the current solid-state battery still cannot compete with state-of-the-art Li-ion technology because of its poor cycle stability and low energy density.

One of the causes of the poor cycle stability in all-solid-state batteries is the lack of chemical stability between the solid electrolytes and electrodes. To efficiently and systematically investigate these chemical compatibility issues, a methodology was developed combining density functional theory calculations and simple experimental techniques such as X-ray diffraction, simultaneous differential scanning calorimetry, thermal gravimetric analysis, and electrochemistry. Two aspects of the chemical stability were investigated: the electrochemical stability of the solid-state conductor, which is relevant wherever the electrolyte contacts an electron pathway, and the electrochemical stability of the electrode/electrolyte interfaces. Computationally derived voltage stability windows are small for both Na3PS4 and Na3PSe4, which are confirmed experimentally. By combining the theory calculations and simple experimental techniques, the most stable system (NaCrO2|Na3PS4|Na–Sn) within a selected Na solid-state battery chemical space (more than 20 different combinations of electrodes and electrolytes) was efficiently found. Important selection criteria for the cathode, electrolyte, and anode in solid-state batteries are derived from this study. This method provides an essential guide for probing the chemical instability issue in the all-solid-state battery, and experimental results show the importance of interfacial engineering for the all-solid-state battery.

The mechanical degradation in all-solid-state batteries can also cause rapid capacity fade. To investigate the severity of the mechanical degradation after long-term cycling, the all-solid-state composite cathode morphology before and after long-term cycling was imaged using the focused ion beam scanning electron microscope tomography. A large amount of void volume (~ 12%) and contact loss area (> 10%) are found in the cathode composite after 50 cycles, revealing the severe mechanical degradation during cycling. The contact loss between the cathode and solid electrolyte is accompanied by the rapid capacity drop during battery cycling. This result suggests that mechanical degradation does not progress gradually starting from the first cycle. Instead, the majority of the void formation and contact loss likely occurs at later cycles and over a short time. This part of the study highlights the importance of mechanical stability for the all-solid-state battery.

Lastly, increasing the cathode active material loading is proposed as a method to increase the full cell energy density of the all-solid-state battery. Current low active material loading in the composite electrode is one of the main reasons for the low full cell energy density. In this part, modeling and experimental results show that a higher cathode loading and therefore an increased energy density can be achieved by increasing the ratio of the cathode to conductor particle size. The utilization of the cathode materials in the solid-state composite is shown to be highly dependent on the particle size ratio of the cathode to the solid electrolyte, especially in the regime of high cathode loading. These results are consistent with the ionic percolation being the limiting factor in cold-pressed solid-state cathode materials and provide specific guidelines on how to improve the energy density of composite cathodes for solid-state batteries. By reducing solid electrolyte particle size and increasing the active material particle size, over 50 vol% cathode loading with high cathode material utilization was experimentally demonstrated. This part of the work proves that a commercially-relevant, energy-dense cathode composite is achievable through simple mixing and pressing method.

Innovations of novel materials often involve synthesizing new compounds with better materials properties. However, computationally designing synthesis methods for these new compounds remains an uncharted new area of research. This thesis proposes to use machine-learning approaches to predict materials synthesis routes by training on synthesis information from the published scientific literature. However, most inorganic materials synthesis information in the scientific literature is locked-up in written natural language and must be parsed using natural language processing and information retrieval techniques. Therefore, this thesis aims to achieve two objectives: 1) constructing a text-mining pipeline that extracts solid-state synthesis datasets from scientific papers, and 2) implementing an interpretable machine-learning method to predict solid-state synthesis conditions.

Training information retrieval systems usually requires large manually labeled datasets, which are not widely available in materials informatics. To alleviate the lack of labeled datasets, we demonstrate a semi-supervised machine-learning method (Chapter 3), which is implemented for the classification of paragraphs in papers. Without any human labeling efforts, latent Dirichlet allocation can cluster keywords into topics corresponding to specific experimental synthesis steps. Guided by a small amount of annotation, supervised training methods, such as random forest, can then associate these steps with different synthesis methods, such as solid-state or hydrothermal synthesis. Using the topic modeling results, we also show a Markov chain representation of the order of experimental steps, which reconstructs a flowchart of synthesis procedures.

To fulfill the first objective, we have extracted a dataset of "codified recipes" for solid-state synthesis using an automated text-mining pipeline (Chapter 4). The dataset currently consists of over 30,000 solid-state synthesis entries. Every entry contains synthesis information including input materials, target materials, experimental operations, the associated processing parameters and synthesis conditions, and the balanced synthesis reaction equation. This dataset is the first-ever collection of machine-readable solid-state synthesis experiments and enables data mining of various aspects of inorganic materials synthesis.

To fulfill the second objective, we have built a machine-learning approach that predicts solid-state synthesis conditions (heating temperature and heating time) using the above-mentioned dataset (Chapter 5). We used dominance importance ranking analysis and discovered that optimal heating temperatures have strong correlations with the stability of precursor materials. This correlation extends Tamman's rule from intermetallics to oxide systems, suggesting the importance of reaction kinetics in solid-state synthesis. Heating times are shown to be strongly correlated with the chosen experimental procedures and instrument setups, which may be indicative of the selection bias in the dataset. Our machine-learning models achieve good synthesis prediction performance and general applicability for diverse chemical systems.

While focusing particularly on solid-state synthesis, this thesis demonstrates a scalable framework to unlock the large amount of inorganic materials synthesis information from the literature, and machine-learn robust and interpretable synthesis predictors. At the end of this thesis, we outline several interesting future research topics which expand the work into a broader context of materials informatics and synthesis science.