RESUMO
A method to predict the crystal structure of equiatomic ternary compositions based only on the constituent elements was developed using cluster resolution feature selection (CR-FS) and support vector machine (SVM) classification. The supervised machine-learning model was first trained with 1037 individual compounds that adopt the most populated ternary 1:1:1 structure types (TiNiSi-, ZrNiAl-, PbFCl-, LiGaGe-, YPtAs-, UGeTe-, and LaPtSi-type) and then validated using an additional 519 compounds. The CR-FS algorithm improves class discrimination and indicates that 113 variables including size, electronegativity, number of valence electrons, and position on the periodic table (group number) influence the structure preference. The final model prediction sensitivity, specificity, and accuracy were 97.3%, 93.9%, and 96.9%, respectively, establishing that this method is capable of reliably predicting the crystal structure given only its composition. The power of CR-FS and SVM classification is further demonstrated by segregating the crystal structure of polymorphs, specifically to examine polymorphism in TiNiSi- and ZrNiAl-type structures. Analyzing 19 compositions that are experimentally reported in both structure types, this machine-learning model correctly identifies, with high confidence (>0.7), the low-temperature polymorph from its high-temperature form. Interestingly, machine learning also reveals that certain compositions cannot be clearly differentiated and lie in a "confused" region (0.3-0.7 confidence), suggesting that both polymorphs may be observed in a single sample at certain experimental conditions. The ensuing synthesis and characterization of TiFeP adopting both TiNiSi- and ZrNiAl-type structures in a single sample, even after long annealing times (3 months), validate the occurrence of the region of structural uncertainty predicted by machine learning.
RESUMO
Two series of Mn-substituted rare-earth zinc arsenides RE(1-y)Mn(x)Zn(2-x)As2 (RE = Eu-Lu) and RE(2-y)Mn(x)Zn(4-x)As4 (RE = La-Nd, Sm, Gd) were prepared by reaction of the elements at 750 °C. Both series are derived from ideal empirical formula REM2As2 (M = Mn, Zn) and adopt crystal structures related to the trigonal CaAl2Si2-type (space group P3m1) in which hexagonal nets of RE atoms and [M2As2] slabs built up of edge-sharing M-centered tetrahedra are alternately stacked along the c-direction. For compounds with divalent RE components (Eu, Yb), the fully stoichiometric and charge-balanced formula REM2As2 is obtained, with Mn and Zn atoms statistically disordered within the same tetrahedral site. For compounds with trivalent RE components, the RE sites become deficient, and the Mn atoms are segregated from the Zn atoms in separate tetrahedral sites. Within the series RE(1-y)Mn(x)Zn(2-x)As2 (Gd-Tm, Lu), the parent CaAl2Si2-type structure is retained, and the Mn atoms are disordered within partially occupied interstitial sites above and below [Zn(2-x)As2] slabs. Within the series RE(2-y)Mn(x)Zn(4-x)As4 (RE = La-Nd, Sm, Gd), the c-axis becomes doubled as a result of partial ordering of Mn atoms between every other pair of [Zn(2-x)As2] slabs. Attempts to synthesize Gd-containing solid solutions with the charge-balanced formula Gd0.67Mn(x)Zn(2-x)As2 suggested that these phases could be formed with up to 50% Mn substitution. Band structure calculations reveal that a hypothetical superstructure model with the formula La1.33MnZn3As4 would have no gap at the Fermi level and that slightly lowering the electron count alleviates antibonding Mn-As interactions; a spin-polarized calculation predicts nearly ferromagnetic half-metallic behavior. X-ray photoelectron spectroscopy confirms the presence of divalent Mn in these compounds.