Determining the crystal structure of solid and liquid materials is important for understanding their mechanical, electromagnetic, and thermodynamic properties. Powder X-ray diffraction (XRD) is an important tool for material characterization, encoding information about crystal symmetry, lattice parameters, types, and the filling of atoms at the nanoscale. However, the current classification method requires a lot of human intervention to complete the classification based on the overall information synthesis evaluation. There are many variables that affect the shape of the XRD pattern, such as the phase or lattice of the material, making it difficult to characterize the material without a known similar structure.
fig. 1 crystal system and space group distributions.
In addition, the presence of some small amounts of impurity phases in the sample can make classification more difficult, time-consuming, and inaccurate. Recent advances in ultrafast synchronous X-ray diffraction and spectroscopy measurements have resulted in the production of extremely large data sets from millions of measurements, far exceeding the amount of data that humans can analyze manually.
fig. 2 diffraction pattern comparison.
Therefore, there is an urgent need for adaptive and automated analysis of XRD data. The performance of the developed deep learning models under different datasets varies greatly, which is characterized by insufficient robustness. Therefore, we need a more robust model that can classify the dynamic and or invisible real XRD data of different materials.
fig. 3 model architectures.
Professor Niaz Abdolrahim's group from the School of Mechanical Engineering at the University of Rochester has developed a generalized deep learning model for crystal systems and space group classification. Since the relative peak intensity, distance, and order in the XRD data characterize the symmetry, the researchers investigated whether there was permutation invariance and translation invariance, and proposed a poolless convolutional neural network (NPCNN) to characterize the material based on relative and local inference between index peaks.
fig. 4 rruff experimental performance.
To achieve a wide range of classification capabilities, the authors have also developed a data generation pipeline to build high-quality datasets that combines experimental effects on diffraction modes and has the ability to simulate materials that have undergone alloying and/or dynamic experiments. In the end, the researchers succeeded in bringing out the state-of-the-art performance of the deep learning model.
fig. 5 rruff and mp dataset confusion matrix. confusion
This study also provides an effective research idea for the development of other spectral characterization technology models. Related**Recently published in NPJ Computational Materials v9: 214 (2023).
fig. 6 materials project performance.
editorial summary
xrd data：an automated deep learning classifier
determining the crystal structure of solid and liquid materials is important for understanding their mechanical, electromagnetic and thermodynamic properties. powder x-ray diffraction (xrd) is an important means of material characterization, encoding information about crystal symmetry, lattice parameters, type, and filling of atoms on nanoscale domains.
fig. 7 lattice augmentation performance.
however, the current classification method requires a lot of human intervention to complete the classification based on comprehensive evaluation of the overall information. there are many variables that affect the shape of an xrd pattern, such as the phase or crystal lattice of the material. without a known similar structure, it is difficult to characterize the material. in addition, the presence of some small amounts of impurity phases in the sample may make classification more difficult and time-consuming. and inaccuracies.
fig. 8 f1 score on rruff and mp datasets.
recent advances in ultrafast synchronized xrd and spectroscopy measurements h**e generated extremely large data sets from millions of measurements, far exceeding what humans can manually analyze. therefore, there is an urgent need for adaptive and automatic analysis of xrd data. the performance of currently developed deep learning models on different data sets varies greatly, showing insufficient robustness. a more robust model is needed that can classify dynamic and/or unseen real xrd data obtained from different materials.
fig. 9 scatterplot on mp performance.
a group led by prof. niaz abdolrahim from the school of mechanical engineering, university of rochester, developed a generalized deep learning model for crystal system and space group classification. considering that the relative peak intensity, distance and order in xrd data indicate symmetry, the researchers investigated whether there is alignment invariance and translation invariance, and based on this, they proposed a no-pool convolutional neural network (npcnn). classification was accomplished by characterizing materials based on relative and local inferences between indexed peaks. to enable extensive classification capabilities, the authors also developed a data generation pipeline to build high-quality data sets that incorporates experimental effects on diffraction patterns. the pipeline also has the capability of simulating materials that undergo alloying and/or dynamic experimentation. the researchers succeeded in **the deep learning model achieve state-of-the-art performance. this study provides a valuable platform for developing models of other spectral characterization techniques. this article was recently published in npj computational materials v9: 214 (2023).
fig. 10 model architecture taxonomy.
Original text abstract and its translation
Automated classification of big X-ray diffraction data using deep learning models
jerardo e. salgado, samuel lerman, zhaotong du, chenliang xu & niaz abdolrahim
abstract in current in situ x-ray diffraction (xrd) techniques, data generation surpasses human analytical capabilities, potentially leading to the loss of insights. automated techniques require human intervention, and lack the performance and adaptability required for material exploration. given the critical need for high-throughput automated xrd pattern analysis, we present a generalized deep learning model to classify a diverse set of materials’ crystal systems and space groups. in our approach, we generate training data with a holistic representation of patterns that emerge from varying experimental conditions and crystal properties. we also employ an expedited learning technique to refine our model’s expertise to experimental conditions. in addition, we optimize model architecture to elicit classification based on bragg’s law and use evaluation data to interpret our model’s decision-**we evaluate our models using experimental data, materials unseen in training, and altered cubic crystals, where we observe state-of-the-art performance and even greater advances in space group classification.
Summary:
In current in-situ X-ray diffraction (XRD) techniques, the ability to generate data exceeds the ability of humans to analyze it, potentially leading to a loss of insight. Automated technology requires human intervention and lacks the performance and adaptability required for materials research. In view of the urgent need for high-throughput automated XRD mode analysis, we propose a generalized deep learning model to classify crystal systems and spatial groups of different materials.
In our approach, we utilize a holistic representation from different experimental conditions and patterns of crystal properties to generate training data. We also employed a rapid learning technique to improve our model's expertise under experimental conditions. In addition, we optimized the model architecture to elicit classifications based on Bragg's law and used the evaluation data to explain our model's decisions. Using experimental data, material not seen in training, and altered cubic crystals, to evaluate the model, we observed state-of-the-art performance and greater advances in spatial group classification.