AI-enhanced discovery and accelerated synthesis of metal phosphosulfides

TL;DR

Metal phosphosulfides offer a rich but underexplored space due to synthesis challenges. The authors combine high-throughput density functional theory with a multi-fidelity transformation to predict experimentally calibrated band gaps from PBEsol data, enabling rapid screening of 909 ternary phosphosulfides and the discovery of 19 thermodynamically stable compounds including Si- and Ge-based compositions. They then extend this approach to experiments via the DADMARS thin-film synthesis workflow, achieving four new phosphosulfide thin films in four combinatorial runs. The integrated theory–AI–experiment pipeline demonstrates viable accelerated discovery for difficult inorganic materials and highlights distinct thiophosphate and non-thiophosphate chemistries governed by the P/S ratio.

Abstract

Metal phosphosulfides have emerged as unique multifunctional materials, but they present unique synthesis challenges compared to more established material classes such as oxides and nitrides. As a consequence, experimental development and theoretical understanding of phosphosulfides have focused on individual compounds rather than on accelerated broad-range exploration. In this work, we first evaluate the synthesizability and band gaps of 909 hypothetical ternary phosphosulfides by density functional theory. We find 19 previously unknown thermodynamically stable compounds, including the first Si- and Ge-based phosphosulfides. For rapid band gap prediction, we then develop a multi-fidelity machine learning model to translate semilocal density functional theory band gaps into experimentally calibrated band gaps. Importantly, we extend the accelerated material development workflow to the experimental domain by demonstrating a route to high-throughput synthesis and characterization of virtually any phosphosulfide material system. The method is based on thin-film combinatorial libraries and yields over 100 unique compositions in each experiment, enabling us to synthesize four distinct phosphosulfide compounds in only four combinatorial experiments without prior synthesis recipes and without compromising on material quality. Thus, we argue that accelerated materials development workflows combining theory, artificial intelligence, synthesis, and characterization can be viable even for experimentally challenging inorganic materials.

Figures (17)

• Figure 1: Summary of the structural and compositional diversity of phosphosulfides. The structures are visualized with VESTA Momma2011. Assignment of oxidation states is discussed in the SI. The "number of materials" (and the corresponding "% of total") refers to the number of phosphosulfides calculated in this study that were found to be both lowest-energy polymorphs and within the 100meV/atom stability tolerance defined in the main text. When these numbers are in bold font, they indicate the three most common structural motifs. A full list of prototypes is given in Table \ref{['tab:mp-id']}. When a composition is labeled as "metallic", it means that it is not charge balanced. The "metal types" are the blocks of the periodic table containing metals that are compatible with the corresponding structural motif. Among d-block metals, d$_0$, d$_{10}$, and d$_\mathrm{os}$ refer to the electronic configuration of the metals after compound formation, with d$_\mathrm{os}$ standing for open-shell configurations.
• Figure 2: Oxidation state of phosphorus as a function of the atomic P/S ratio in the ternary phosphosulfides calculated in this study (only lowest-energy polymorphs within stability tolerance). The P/S ratios are discrete and are linked to the generalized compositions shown on the top x axis. The data points are jittered for a more intuitive visualization of the number of materials at each P/S ratio.
• Figure 3: Number of synthesized phosphosulfides as a function of their energy above the stability hull. Only the lowest-energy polymorphs found in study are included in the statistics. The percentage of synthesized materials for each $E_h$ range is indicated.
• Figure 4: Energy above hull heat maps of all the lowest-energy phosphosulfide polymorphs calculated in this study. Each M$^{x+}$ heat map includes phosphosulfides of metals with assumed oxidation state $x$. Along the x axes, materials with different anion motifs (top) are presented, which correspond to specific compositions (bottom). The motifs are visualized in Fig. \ref{['fig:motifs']}. The color of the materials with $E_h = 100meV/atom$ is also used for materials with higher energies above hull. White cells indicate that the corresponding material has not been calculated. Cells marked with an "s" indicate that the material is present in the ICSD database and is therefore identified as synthesized. The dashed vertical lines separate thiophosphate compositions ("Thio") from non-thiophosphate compositions ("Non-Thio"). The lowest-energy prototypes of each material are shown in Fig. \ref{['fig:HM_Versions']}.
• Figure 5: Ag-P-S and Cu-P-S ternary systems explored by high-throughput experiments. The data points correspond to the compositions obtained from combinatorial synthesis. The three axes indicate the atomic composition of each of the three elements, normalized to a sum of 100%. Depending on synthesis conditions such as temperature, these data points may be single-phase compounds (e.g., Ag3PS4) like the computationally screened ones, but also multi-phase systems (e.g., a mix of Ag2S and Ag7PS6) or solid solutions (e.g. Ag_2+3x(P_xS)). Single-phase compounds synthesized in this study are marked as blue circles. Other compounds calculated in this study are marked as red circles.