Abstract:Understanding structure-property relationships is essential for advancing technologies based on thin films. X-ray pair distribution function (PDF) analysis can access relevant atomic structure details spanning local-, mid-, and long-range order. While X-ray PDF has been adapted for thin films on amorphous substrates, measurements on single crystal substrates are necessary to accurately determine structure origins for some thin film materials, especially those for which the substrate changes the accessible structure and properties. However, when measuring films on single crystal substrates, high intensity anisotropic Bragg spots saturate 2D detector images, overshadowing the thin films' isotropic scattering signal. This renders previous data processing methods for films on amorphous substrates unsuitable for films on single crystal substrates. To address this measurement need, we developed IsoDAT2D, an innovative data processing approach using unsupervised machine learning algorithms. The program combines non-negative matrix factorization and hierarchical agglomerative clustering to separate thin film and single crystal substrate X-ray scattering signals. We use SimDAT2D, a program we developed to generate synthetic thin film data, to validate IsoDAT2D. We also use IsoDAT2D to isolate X-ray total scattering signal from a thin film on a single crystal substrate. The resulting PDF data are compared to similar data processed using previous methods, demonstrating superior performance relative to substrate subtraction with a single crystal substrate and similar performance to substrate subtraction from an amorphous substrate. With IsoDAT2D, there are new opportunities to expand PDF to a wider variety of thin films, including those on single crystal substrates, with which new structure-property relationships can be elucidated to enable fundamental understanding and technological advances.

What problem does this paper attempt to address?

### Problems Addressed by the Paper This paper aims to address the issue of separating X-ray total scattering signals when measuring thin film materials on single crystal substrates. Specifically, when measuring thin film materials on single crystal substrates, high-intensity anisotropic Bragg spots saturate the 2D detector images, obscuring the isotropic scattering signals of the thin film. This makes traditional data processing methods (such as subtracting the scattering signal from a bare substrate) ineffective for thin film materials on single crystal substrates. ### Background and Challenges 1. **Importance of Thin Film Materials**: - Thin film materials have unique properties in applications such as spintronics, phase-change memory, photonics, and semiconductor devices, which differ from their bulk counterparts. - With the increasing demand for high-performance computers and advanced electronic technologies, there is a need to continuously improve the performance of these materials. 2. **Structure-Property Relationship**: - Understanding the structure-property relationship of materials is crucial for advancing technologies based on thin film materials. - The choice of single crystal substrate affects the structure and properties of the thin film, thereby influencing its performance. 3. **Limitations of Existing Technologies**: - X-ray diffraction (XRD) is primarily used to determine the atomic structure of crystalline thin films, but for amorphous or disordered materials, traditional crystallographic methods (such as XRD) cannot capture local structural features. - For thin films on single crystal substrates, traditional substrate subtraction methods are ineffective because the strong Bragg spots lead to residual signals and over-subtraction, resulting in inaccurate PDF data. ### Solution To overcome this challenge, the authors developed a new data processing method called IsoDAT2D. This method combines non-negative matrix factorization (NMF) and hierarchical clustering algorithm (HAC) to effectively separate the X-ray scattering signals of the thin film and the single crystal substrate. The specific steps are as follows: 1. **Rotation and Integration**: - Rotate and integrate the 2D scattering images to generate multiple 1D datasets for machine learning algorithms to process. 2. **Non-Negative Matrix Factorization (NMF)**: - Use NMF to identify recurring components corresponding to the scattering signals of the thin film and the substrate. 3. **Hierarchical Clustering (HAC)**: - Use HAC to cluster the components generated by NMF, grouping similar components to separate the isotropic scattering signal of the thin film. 4. **Averaging and Smoothing**: - Average and smooth the components in the selected clusters to generate high-quality 1D X-ray scattering data, from which PDF data can be derived. ### Experimental Validation The authors used the SimDAT2D program to generate synthetic data to validate the effectiveness of IsoDAT2D and applied it to experimental data, successfully separating the isotropic scattering signal from a Ge2Sb2Te5 thin film on a single crystal Si substrate. The results showed that the quality of the PDF data generated using IsoDAT2D was significantly better than that obtained using traditional substrate subtraction methods. ### Conclusion By developing and applying IsoDAT2D, researchers can more effectively separate the X-ray total scattering signals of thin film materials on single crystal substrates, thereby better understanding the structure-property relationship of thin films and advancing related technologies.

Distinguishing Isotropic and Anisotropic Signals for X-ray Total Scattering using Machine Learning

Data-driven approach for synchrotron X-ray Laue microdiffraction scan analysis

Artifact Identification in X-ray Diffraction Data using Machine Learning Methods

Artifact identification in X‐ray diffraction data using machine learning methods

Deconvoluting Thermomechanical Effects in X-ray Diffraction Data using Machine Learning

Harnessing interpretable and unsupervised machine learning to address big data from modern X-ray diffraction

Integrated analysis of X-ray diffraction patterns and pair distribution functions for machine-learned phase identification

A machine learning approach to map crystal orientation by optical microscopy

Exploring Supervised Machine Learning for Multi-Phase Identification and Quantification from Powder X-Ray Diffraction Spectra

Separating edges from microstructure in X-ray dark-field imaging: Evolving and devolving perspectives via the X-ray Fokker-Planck equation

Fast and interpretable classification of small X-ray diffraction datasets using data augmentation and deep neural networks

Machine Learning in X-ray Scattering for Materials Discovery and Characterization

Diffractive small angle X-ray scattering imaging for anisotropic structures

Promoting X‐ray scattering data analysis with two‐dimensional correlation spectroscopy

Advantages of a curved image plate for rapid laboratory-based x-ray total scattering measurements: Application to pair distribution function analysis.

Interpretable structure-property correlation in X-Ray diffraction patterns of HfZrO thin films via machine learning

Towards scanning nanostructure X‐ray microscopy

An Automatic Regularization Method: An Application for 3-D X-Ray Micro-CT Reconstruction Using Sparse Data

Analysis of extended X-ray absorption fine structure (EXAFS) data using artificial intelligence techniques