Geoffroy Hautier

Zintl phosphide BaCd2P2 (BCP) has recently received attention as a promising impurity- and defect-tolerant solar absorber. It has been found to exhibit bright band-to-band photoluminescence (PL) comparable to that of GaAs. In this work, we systematically investigate the mechanisms of radiative and nonradiative recombination in BCP by studying the band-to-band and defect PL emissions. From the temperature quenching of the band-to-band PL, we find that BCP exhibits an apparent activation energy for nonradiative recombination similar to that of GaAs. Furthermore, both materials possess radiative shallow defects with PL emissions near their respective bandgaps. In both cases, these defect emissions arise from free-to-bound transitions. By combining first-principles calculations and experiments, we find that the shallow radiative-defect emissions are most consistent with Ba vacancies (VBa) in BCP and B-on-As substitutional defects (BAs) in GaAs. The former is supported by the suppression of shallow-defect PL emission in Ba-rich BCP samples. Finally, we discuss emissions from deep defects at approximately 0.95 eV at low temperatures, which are present in both materials.

Point defects play a central role in driving the properties of materials. First-principles methods are widely used to compute defect energetics and structures, including at scale for high-throughput defect databases. However, these methods are computationally expensive, making machine learning force fields (MLFFs) an attractive alternative for accelerating structural relaxations. Most existing MLFFs are based on graph neural networks (GNNs), which can suffer from oversmoothing, oversquashing, and poor representation of long-range interactions. Both of these issues are especially of concern when modeling point defects. To address these challenges, we introduce the Accelerated Deep Atomic Potential Transformer (ADAPT), an MLFF that replaces graph representations with a direct coordinates-in-space formulation and explicitly considers all pairwise atomic interactions. Atoms are treated as “tokens,” with a transformer encoder modeling their interactions. Applied to a dataset of silicon point defects, ADAPT achieves an ∼ 22 % reduction in force and an ∼ 40 reduction in energy prediction error relative to a state-of-the-art GNN-based model, while requiring only a fraction of the computational cost.

Phonons play an essential role in condensed matter physics, influencing key phenomena such as vibrational entropy, thermal conductivity, superconductivity, ferroelectricity, and photoluminescence spectra. Here, we present a comprehensive database of harmonic phonon properties, constructed using automated high-throughput (HT) density functional theory (DFT) calculations for 26,413 compounds. The database covers materials with all seven crystal systems and includes primitive cells with up to 60 atoms. In this work, phonons were computed using DFT-based (PBEsol level of theory) second-order interatomic force constants (IFCs), obtained from perturbed supercell calculations and fitted using either least-squares or LASSO-based regression depending on the number of required finite displacements. The phonon workflow is implemented in the HT software Atomate2, incorporating Pheasy, a compressive sensing lattice dynamics code. Within this framework, space group and point group symmetries, as well as the acoustic sum rule and rotational invariance constraints, are applied to the force constants to ensure physical accuracy and reduce numerical errors. This approach offers a substantial computational speedup compared to both the traditional finitedisplacement method and density functional perturbation theory, while maintaining accuracy comparable to both. The resulting phonon database includes phonon dispersions, phonon density of states (DOS), and derived thermodynamic properties such as Helmholtz free energy ( F ), entropy ( S ), and constant-volume heat capacity ( C _V ). Our work not only establishes an HT methodology for phonon calculations, but also delivers a large-scale phonon database accessible to Materials Project (MP) users for a range of applications, including materials screening and follow-up computational studies.

The AM ₂ Pn ₂ Zintl compounds are a large class of semiconductor materials that have a wide range of bandgaps and are mostly stable in the same crystal structure. Representative compounds BaCd ₂ P ₂ and CaZn ₂ P ₂ have recently been found to exhibit high visible light absorption and long carrier lifetime. Here we use high throughput first-principles calculations to study AM ₂ Pn ₂ alloys for applications as tandem top cell absorbers (i.e. bandgaps around 1.8 eV) and far infrared (IR) detector materials (i.e. bandgaps lower than 0.5 eV). Using a first-principles computational screening workflow for assessing stability and electronic structure of alloys, we identify several promising candidates. These include Ca(Cd _0.8 Mg _0.2 ) ₂ P ₂ with a suitable direct bandgap for use in tandem top cells on silicon bottom cells and SrCd ₂ (Sb _{1− x} Bi _x ) ₂ for far IR detectors. We demonstrate that alloys of AM ₂ Pn ₂ materials can be realized by experimentally synthesizing Ca(Zn _0.8 Mg _0.2 )2P ₂ .

Zintl compounds of the Ca₁₄AlSb₁₁ structure type, particularly Yb₁₄MSb₁₁ (M = Mg, Mn, Zn), have emerged as leading contenders for high-temperature thermoelectric applications due to their remarkably low thermal conductivity and excellent electronic properties. In contrast, lighter arsenic-containing compounds have received less attention, as they are typically expected to have poor thermoelectric performance due to higher thermal conductivity. In this study, we have demonstrated that nonpropagating heat conduction, known as diffuson, dominates lattice thermal conductivity in Zintl arsenides of the Ca₁₄AlSb₁₁ structure type, resulting in exceptionally low lattice thermal conductivity. We have introduced a new member of this family, Eu₁₄MgAs₁₁, synthesized using metal hydride and binary precursor-based methods through ball milling and high-temperature annealing. Additionally, we have synthesized and measured the transport properties of Eu₁₄ZnAs₁₁, Eu₁₄CdAs₁₁, and Eu₁₄Zn_1.2As₁₁ to illustrate that diffuson conduction prevails across these compositions, leading to lattice thermal conductivity of ∼0.55 W m^-1 K^-1 at 298 K and defying the conventional correlation between mass and lattice thermal conductivity. Moreover, the band structures of these compounds exhibit multivalley bands near the Fermi level, resulting in a Seebeck coefficient of 226 μV/K at 1252 K for the Mg analog. The combination of low thermal conductivity and high valley degeneracy allows these compounds to achieve impressive thermoelectric efficiency, with Eu₁₄MgAs₁₁ reaching a maximum zT of 1.3 at 1250 K. Furthermore, these findings indicate that the traditional understanding that lighter compounds exhibit high thermal conductivity does not apply to such complex crystal structures, suggesting a new direction for future thermoelectric research.