Pyrochlore compound (A₂B₂X₇) is one of the precious crystal materials that are important for energy innovation. It is applicable as ionic conductors for solid state batteries, environment catalyst for clean energy production and as bioanode in fuel cell of microbial. Accurate determination of lattice constant would help in materials characterization. Example of lattice constant location of Lu₂V₂O₇ pyrochlore on the cubic structure is shown on Figure 1. The accurate knowledge of the lattice parameter would give information about the compound thermal properties, strain state and defect in the structure. Due to that, it is important for precise lattice constant value to be determined. Most of the techniques to determine the lattice parameter is by using X-Ray Diffraction which is tedious and costly. To overcome the problem, we proposed an accurate machine learning algorithm to model the pyrochlore compound lattice value.

In our research [1], we proposed an optimized Machine Learning (ML) model based on Particle Swarm Optimization paired with Support Vector Regression (SVR) algorithm. Our input data of 220 dataset to predict the lattice constant are based on ionic radii (rA, rB and rX) and electronegativities (xA, xB and xX) values of each cation (A and B) and anion (X). We validate our results using multiple fold technique and found that 15-fold method achieve the best results as shown in Figure 2. Predicted value using kernel SVR function of Radial Basis Function outperformed the others as the average value of Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Correlation Coefficient (CC) is evaluated at 0.0122, 0.0099, and 0.9990, respectively. We checked our results with previously reported findings using other ML model such as Bayesian SVR [2], Artificial Neural Network [2], and statistical algorithm using linear model [3] to confirm the superiority of our techniques. The RMSE and CC improve 64% and 1% respectively compared to the second-best model. Through this analysis, the modelling based on ML technique can provide a highly accurate prediction in accessing the pyrochlore compounds crystal structure properties which means it could replace the traditional measurement technique effectively.

 Figure 1: Lattice constant and other parameters for cubicstructure of Lu₂V₂O₇ pyrochlore compound.

Figure 2: Comparison of predicted PSO-SVR model with experimental values using 15-fold validation technique.

References

[1] Mohamad Zamri, I.U., Abd Rahman, M.A., Anak Bundak, C.E., “Prediction of Lattice Constant of Pyrochlore Compounds Using Optimized Machine Learning Model”, (2023) ETTIS 2023, Noida, India.

[2] Alade, I.O., Oyedeji, M.O., Abd Rahman, M.A., Saleh, T.A.: Prediction of the lattice constants of pyrochlore compounds using machine learning. Soft computing. 26, 8307–8315 (2022). https://doi.org/10.1007/s00500-022-07218-1.

[3] Brik, M.G., Srivastava, A.M.: Pyrochlore structural chemistry: Predicting the lattice constant by the ionic radii and electronegativities of the constituting ions. Journal of the American Ceramic Society. 95, 1454–1460 (2012). https://doi.org/10.1111/j.1551-2916.2012.05115.x.

About Author

Mohd Amiruddin Abd Rahman (Assoc. Prof. Dr.) Physics Department, Faculty of Science, Universiti Putra Malaysia Kepakaran: Indoor Positioning, Signal Processing, Internet of Things and & UAV, Deep and Machine Learning, Artificial Intelligence, Big Data Analytics Email: mohdamir@upm.edu.my

Date of Input: 30/05/2024 | Updated: 09/07/2024 | harithdaniel