Random forest classifies multi-rare-earth disilicate phase compositions and yields mean-and-deviation radius criteria for single-phase beta/gamma design
Synopsis
This work uses a random forest model to classify the phase composition of multi-rare-earth-principal-component RE2Si2O7 disilicates into single-beta, single-gamma, single-delta/mixed delta+gamma, and separate phase, identifies the average RE3+ cationic radius and the deviation of RE3+ cationic radius as the most influential factors, validates the model by predicting the phase compositions of (Gdx1Hox2Ybx3Lux4)2Si2O7 and (Ndx1Hox2Ybx3Lux4)2Si2O7 systems with experimental characterization of representative compositions, links phase formation through high-throughput DFT calculations to low energy costs for accommodating configurational randomness and rapid convergence of the configurational entropy of mixing with increased excitation energy, and establishes quantitative design criteria for si
Interpretation
A high-accuracy random forest classification model is developed that sorts multi-rare-earth-principal-component RE2Si2O7 phase compositions into single-beta, single-gamma, single-delta/mixed delta+gamma, and separate phase, identifying the average RE3+ cationic radius and the deviation of RE3+ cationic radius as the most influential factors. Previously the phase selection of multi-rare-earth-principal-component disilicates was difficult to predict because it varies with the elemental properties of the RE cationic sites; this work turns phase-composition prediction into a classification problem keyed on average radius and radius deviation. The abstract reports high accuracy and names the two dominant factors; specific accuracy values, dataset size, and feature-engineering details are not given in the abstract.
The trained model predicts the phase compositions of the (Gdx1Hox2Ybx3Lux4)2Si2O7 and (Ndx1Hox2Ybx3Lux4)2Si2O7 systems, supported by experimental characterization of representative compositions. The model prediction is extended from the training space to two quaternary rare-earth-principal-component systems and checked against experiments on representative compositions rather than stopping at statistical fitting. The abstract explicitly mentions prediction and experimental characterization of representative compositions, but gives no experimental methods, sample counts, or numerical characterization results.
High-throughput DFT calculations show that phase formation correlates with low energy costs to accommodate configurational randomness in the multicomponent system, characterized by rapid convergence of the configurational entropy of mixing with increased excitation energy. Phase stability is extended beyond a purely cationic-radius descriptor to an energetic view of the cost of configurational disorder and the convergence behavior of the configurational entropy of mixing. The abstract reports high-throughput DFT calculations and the qualitative convergence of configurational entropy, without specific energy values or calculation scale.
Quantitative design criteria for single-phase beta-(nRExi)2Si2O7 and gamma-(nRExi)2Si2O7 are established: beta polymorphs require an average radius below 0.885 A, gamma polymorphs require an average radius between 0.885 A and 0.900 A, and the radius deviation must be sufficiently small, with an upper bound that increases monotonically with the average radius and reaches about 0.04 near 0.885 A and 0.900 A. Phase selection in multi-rare-earth-principal-component disilicates is advanced from empirical trial and error to an actionable composition design window aimed directly at environmental barrier coating candidate screening. The criteria are supported jointly by the random forest model and DFT calculations, with experimental validation on representative compositions; the abstract does not give error ranges or an applicable temperature window for the criteria.
Perspective
The result targets multi-rare-earth-principal-component RE2Si2O7 disilicates that must retain a stable beta or gamma polymorph under high-temperature service conditions, serving composition screening for environmental barrier coating candidates; the design criteria take the average RE3+ cationic radius and the radius deviation as inputs, apply to phase-selection settings dominated by these two descriptors, and can support prediction and representative-composition validation for quaternary rare-earth-principal-component systems such as (Gdx1Hox2Ybx3Lux4)2Si2O7 and (Ndx1Hox2Ybx3Lux4)2Si2O7.
The visible text is only the abstract, lacking figures, accuracy metrics, dataset size, DFT calculation parameters, and experimental characterization details, so the error range of the criteria, the applicable temperature window, and performance across broader rare-earth combinations remain open questions; the relation in which the upper bound of the radius deviation increases monotonically with the average radius and reaches about 0.04 near 0.885 A and 0.900 A also needs confirmation of its boundary behavior in the full data.