TY - JOUR

T1 - Identification of impurity configurations in quantum dots from photoionization spectra using machine learning

AU - Nahapetyan, G.L.

AU - Kolotyan, G.

AU - Zeiri, Nabil

AU - Резник, Родион Романович

AU - Bessarab, P.F.

AU - Mantashyan, Paytsar

AU - Hayrapetyan, David

PY - 2026/10/1

Y1 - 2026/10/1

N2 - Determining the structural characteristics of impurity-doped quantum dots from spectroscopic measurements represents a challenging inverse problem in semiconductor nanostructure research. In this work, we investigate the feasibility of combining theoretical spectroscopy with machine learning techniques for inverse identification and automated characterization of semiconductor quantum nanostructures. First, the direct problem is solved for a strongly oblate ellipsoidal quantum dot containing multiple hydrogen-like donor impurities using the finite element method. Several impurity arrangements, including linear, central, circular, asymmetric, sinusoidal, and crosswise configurations, are considered for different numbers of impurities and quantum dot sizes. For each configuration, the electron probability density, energy spectrum, and corresponding photoionization cross-section spectra are calculated, forming a synthetic dataset for the inverse analysis. The resulting spectra are then used to train and evaluate two machine learning classifiers: Support Vector Machine (SVM) and a Multi-Layer Perceptron (MLP). The SVM model demonstrates the best performance, achieving a classification accuracy of approximately 94%, while the MLP reaches about 90%. These results indicate that impurity-induced modifications of the photoionization spectra provide distinct spectral fingerprints that enable reliable identification of impurity configurations.

AB - Determining the structural characteristics of impurity-doped quantum dots from spectroscopic measurements represents a challenging inverse problem in semiconductor nanostructure research. In this work, we investigate the feasibility of combining theoretical spectroscopy with machine learning techniques for inverse identification and automated characterization of semiconductor quantum nanostructures. First, the direct problem is solved for a strongly oblate ellipsoidal quantum dot containing multiple hydrogen-like donor impurities using the finite element method. Several impurity arrangements, including linear, central, circular, asymmetric, sinusoidal, and crosswise configurations, are considered for different numbers of impurities and quantum dot sizes. For each configuration, the electron probability density, energy spectrum, and corresponding photoionization cross-section spectra are calculated, forming a synthetic dataset for the inverse analysis. The resulting spectra are then used to train and evaluate two machine learning classifiers: Support Vector Machine (SVM) and a Multi-Layer Perceptron (MLP). The SVM model demonstrates the best performance, achieving a classification accuracy of approximately 94%, while the MLP reaches about 90%. These results indicate that impurity-induced modifications of the photoionization spectra provide distinct spectral fingerprints that enable reliable identification of impurity configurations.

KW - Impurity configurations

KW - Machine learning

KW - Photoionization cross-section

KW - Quantum dots

KW - Spectroscopic fingerprinting

UR - https://www.mendeley.com/catalogue/1fe3291e-1c89-3305-883f-5c677cada49d/

U2 - 10.1016/j.ssc.2026.116551

DO - 10.1016/j.ssc.2026.116551

M3 - Article

VL - 417

JO - Solid State Communications

JF - Solid State Communications

SN - 0038-1098

M1 - 116551

ER -