In this paper, we demonstrate that digital technologies can be successfully applied to image analysis and prediction of the properties of functional materials using the example of a new method for rapidly identifying crystallographic orientation parameters of multicrystalline silicon. The proposed method is based on machine learning. The analysis of textured multicrystalline silicon wafers is carried out using the original single-crystal grain clustering algorithm, and the crystallographic orientation parameters are identified using a neural network model. The identification is based on the correlation of the contrast of the macrostructure display associated with the reflective features of the grains and their orientation parameters. Neural network architecture—multilayer perceptron—is chosen taking into account the restrictions on the amount of input data. However, in conjunction with the algorithm, the optimal amount of training data satisfies the requirements of the neural network training process and ensures high efficiency in identifying orientation parameters on scanned images of textured multicrystalline silicon wafers. Together with the algorithm, the optimal amount of training data satisfies the requirements of the neural network training process and ensures high efficiency in identifying orientation parameters on scanned images of textured multicrystalline silicon wafers.

In this paper, we analyze the contrast of images obtained by scanning electron microscopy (SEM) in the backscattered electron mode, which shows that the contrast on the polished surface of multicrystalline silicon (multisilicon) is due to the parameters of misorientation between grains. The contrast values corresponding to special oblique boundaries with various inverse density values of the coinciding nodes of the crystal lattices of neighboring grains forming these boundaries are established. Since the SEM method allows relatively large surface areas of the samples (~1 cm2) to be studied, on account of the method proposed here for identifying special oblique boundaries, it is possible to analyze the nature of the interaction of boundaries and their distribution density in the structure, depending on certain crystallization conditions. In SEM studies of multisilicon, this can be important for initial analysis of the structure and further formulation of research problems without the use of special selective etchants that destroy the surface of the studied samples.

The geometric parameters of the grain boundaries that make up triple joints and their possible combinations are studied by electron backscatter diffraction using Channel HKL5 software. It is shown experimentally that, in addition to the previously studied and fairly widespread triple joints of special grain boundaries, there are other varieties of them. It is found that at these triple junctions, in particular, the geometric combination rule is violated, which may indirectly indicate the recombination activity of special boundaries.

The orientation of grains and the special boundaries formed by them in multicrystalline silicon has been studied by electron backscattered diffraction. It is found that the crystallographic parameters of special boundaries (misorientation angle and rotation axis) obtained using the Tango HKL Channel 5 software module may differ from the results of their direct computation by calculating the rotation matrix in the Spyder integrated development environment, using identical formulas based on Euler angles. In particular, special boundaries ∑3, ∑9, and ∑27a with misorientation angles of 180°, 120°, and 165°, respectively, are found in multicrystalline silicon. These versions of special grain boundaries are theoretically possible for the crystals of cubic system; however, they have not been investigated experimentally in multicrystalline silicon.