Application of Computational Intelligence to Investigation of Defect Centers in Semi-Insulating Materials by Photoinduce
- PDF / 210,960 Bytes
- 6 Pages / 612 x 792 pts (letter) Page_size
- 104 Downloads / 179 Views
0994-F03-14
Application of Computational Intelligence to Investigation of Defect Centers in SemiInsulating Materials by Photoinduced Transient Spectroscopy Pawel Kaminski1, Stanislaw Jankowski2, Roman Kozlowski1, and Janusz Bedkowski2 1 Epitaxy Department, Institute of Electronic Materials Technology, Wolczynska 133, Warszawa, 01-919, Poland 2 Institute of Electronic Systems, Warsaw University of Technology, Nowowiejska 15/19, Warszawa, 00-665, Poland ABSTRACT A computational intelligence algorithm has been applied to extracting trap parameters from the photocurrent relaxation waveforms recorded at the temperature range of 20-320 K for semiinsulating (SI) InP samples. Using the inverse Laplace transform procedure, the spectral surfaces, visualized in the three dimensional space as functions of temperature and emission rate, are calculated. The processes of thermal emission of charge carriers from defect centers manifest themselves as the sharp folds on the spectral surface. Using a set of Gaussian functions, the approximating surface is created and the ridgelines of the folds, giving the temperature dependences of the emission rate for the detected traps, are determined. The approximation is performed using the support vector machine (SVM) algorithm that allows for trading off between the model complexity and fitting accuracy. The new approach is exemplified by comparing the trap properties for SI InP wafers after annealing in iron phosphide and pure phosphorous atmospheres. INTRODUCTION Obtaining new semiconductor materials and manufacturing new devices involve a new approach to characterization of material and device parameters. The modern metrology paradigm requires fast processing of large experimental datasets [1]. To reduce the time of the computeraided analysis and extraction of the parameters, there is a necessity to apply computational intelligence algorithms including neural networks (NN), support vector machines (SVM) and other kernel machines [2, 3]. In this paper the SVM algorithm has been implemented to characterization of defect levels in semi-insulating (SI) InP. The quality assessment of SI InP wafers is of great importance in terms of manufacturing high frequency devices and optoelectronic integrated circuits for highspeed telecommunication and computer networks. The defect levels strongly affect the material properties. So far, a high resistivity of InP, ranging from 107-109 Ωcm, has been mainly achieved by doping with iron the InP crystals grown by a high-pressure liquid encapsulated Czochralski (LEC) method. Iron atoms act as deep acceptors compensating residual shallow donors resulting from contamination with silicon and sulphur atoms [4]. On the other hand, the attempts have been made recently to obtain the material semi-insulating properties by annealing the undoped SI InP wafers in iron-phosphide (IP) or pure phosphorus (PP) atmosphere [5]. The photoinduced transient spectroscopy (PITS) is a very effective tool for investigation of defect centers in high-resistivity semiconductors [1, 5]. T
Data Loading...
