With the goal of finding new lithium solid electrolytes by a combined computational–experimental method, the exploration of the Li–Al–O–S phase field resulted in the discovery of a new sulfide Li3AlS3. The structure of the new phase was determined through an approach combining synchrotron X-ray and neutron diffraction with 6Li and 27Al magic-angle spinning nuclear magnetic resonance spectroscopy and revealed to be a highly ordered cationic polyhedral network within a sulfide anion hcp-type sublattice. The originality of the structure relies on the presence of Al2S6 repeating dimer units consisting of two edge-shared Al tetrahedra. We find that, in this structure type consisting of alternating tetrahedral layers with Li-only polyhedra layers, the formation of these dimers is constrained by the Al/S ratio of 1/3. Moreover, by comparing this structure to similar phases such as Li5AlS4 and Li4.4Al0.2Ge0.3S4 ((Al + Ge)/S = 1/4), we discovered that the AlS4 dimers not only influence atomic displacements and Li polyhedral distortions but also determine the overall Li polyhedral arrangement within the hcp lattice, leading to the presence of highly ordered vacancies in both the tetrahedral and Li-only layer. AC impedance measurements revealed a low lithium mobility, which is strongly impacted by the presence of ordered vacancies. Finally, a composition–structure–property relationship understanding was developed to explain the extent of lithium mobility in this structure type.

The selection of the elements to combine delimits the possible outcomes of synthetic chemistry because it determines the range of compositions and structures, and thus properties, that can arise. For example, in the solid state, the elemental components of a phase field will determine the likelihood of finding a new crystalline material. Researchers make these choices based on their understanding of chemical structure and bonding. Extensive data are available on those element combinations that produce synthetically isolable materials, but it is difficult to assimilate the scale of this information to guide selection from the diversity of potential new chemistries. Here, we show that unsupervised machine learning captures the complex patterns of similarity between element combinations that afford reported crystalline inorganic materials. This model guides prioritisation of quaternary phase fields containing two anions for synthetic exploration to identify lithium solid electrolytes in a collaborative workflow that leads to the discovery of Li3.3SnS3.3Cl0.7. The interstitial site occupancy combination in this defect stuffed wurtzite enables a low-barrier ion transport pathway in hexagonal close-packing.

Mixed anion materials and anion doping are very promising strategies to improve solid-state electrolyte properties by enabling an optimized balance between good electrochemical stability and high ionic conductivity. In this work, we present the discovery of a novel lithium aluminum sulfide–chloride phase, obtained by substitution of chloride for sulfur in Li3AlS3 and Li5AlS4 materials. The structure is strongly affected by the presence of chloride anions on the sulfur site, as the substitution was shown to be directly responsible for the stabilization of a higher symmetry phase presenting a large degree of cationic site disorder, as well as disordered octahedral lithium vacancies. The effect of disorder on the lithium conductivity properties was assessed by a combined experimental–theoretical approach. In particular, the conductivity is increased by a factor 103 compared to the pure sulfide phase. Although it remains moderate (10–6 S·cm–1), ab initio molecular dynamics and maximum entropy (applied to neutron diffraction data) methods show that disorder leads to a 3D diffusion pathway, where Li atoms move thanks to a concerted mechanism. An understanding of the structure–property relationships is developed to determine the limiting factor governing lithium ion conductivity. This analysis, added to the strong step forward obtained in the determination of the dimensionality of diffusion, paves the way for accessing even higher conductivity in materials comprising an hcp anion arrangement.

A hexagonal analogue, Li 6 SiO 4 Cl 2 , of the cubic lithium argyrodite family of solid electrolytes is isolated by a computation–experiment approach. We show that the argyrodite structure is equivalent to the cubic antiperovskite solid electrolyte structure through anion site and vacancy ordering within a cubic stacking of two close-packed layers. Construction of models that assemble these layers with the combination of hexagonal and cubic stacking motifs, both well known in the large family of perovskite structural variants, followed by energy minimization identifies Li 6 SiO 4 Cl 2 as a stable candidate composition. Synthesis and structure determination demonstrate that the material adopts the predicted lithium site-ordered structure with a low lithium conductivity of ∼10 –10 S cm –1 at room temperature and the predicted hexagonal argyrodite structure above an order–disorder transition at 469.3(1) K. This transition establishes dynamic Li site disorder analogous to that of cubic argyrodite solid electrolytes in hexagonal argyrodite Li 6 SiO 4 Cl 2 and increases Li-ion mobility observed via NMR and AC impedance spectroscopy. The compositional flexibility of both argyrodite and perovskite alongside this newly established structural connection, which enables the use of hexagonal and cubic stacking motifs, identifies a wealth of unexplored chemistry significant to the field of solid electrolytes.

The application of machine learning models to predict material properties is determined by the availability of high-quality data. We present an expert-curated dataset of lithium ion conductors and associated lithium ion conductivities measured by a.c. impedance spectroscopy. This dataset has 820 entries collected from 214 sources; entries contain a chemical composition, an expert-assigned structural label, and ionic conductivity at a specific temperature (from 5 to 873 °C). There are 403 unique chemical compositions with an associated ionic conductivity near room temperature (15–35 °C). The materials contained in this dataset are placed in the context of compounds reported in the Inorganic Crystal Structure Database with unsupervised machine learning and the Element Movers Distance. This dataset is used to train a CrabNet-based classifier to estimate whether a chemical composition has high or low ionic conductivity. This classifier is a practical tool to aid experimentalists in prioritizing candidates for further investigation as lithium ion conductors.