Computer-assisted infrared spectra interpretation for amorphous silicon alloys

https://doi.org/10.1016/j.jqsrt.2004.12.028Get rights and content

Abstract

A computer program for the structural interpretation of the infrared (IR) spectra is developed and tested. The interpretation of the IR spectra is made by using an hybrid system which includes library search and rule-based interpretation methods together.

The computer programs were written in Pascal Codes. The prototype IR library of silicon alloys includes amorphous silicon (a-Si), amorphous silicon dioxide (a-SiOx), amorphous silicon nitride (a-Si3N4) and amorphous silicon carbide (a-SiC) references. The known spectra of these compounds were fed into the system as an unknown samples. The performance of the developed program was evaluated on a test set of 157 spectra and the percentages of successful identification ranged between 78% and 99% for different alloys.

Introduction

Amorphous silicon (a-Si) alloys are important materials for a-Si-based devices. These alloys are widely used in the electronic applications such as solar cells, phototransistors, image sensors, light emitting devices and so on. On the other hand, during the past decades Fourier transform infrared (FTIR) spectroscopy has become one of the most effective and commonly used analytical techniques for qualitative and quantitative analysis of these alloys. FTIR spectroscopy can be also used to obtain doping levels in deposited amorphous and polycrystalline semiconductor thin film. In these studies computer-assisted qualitative analysis were presented. Even though there are preliminary results showing that this method can be applied to the quantitative analysis of a-SiC thin films too, but they are not presented in this paper.

FTIR spectroscopy are commonly used for the structural characterization of amorphous semiconductor alloys [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20]. In this study, only four of amorphous silicon alloys, (a-SiC, a-SiN, a-Si, a-SiOx), were considered but it can be developed for larger groups. In this group SiC is the most hardly identified compound, it can be misclassified as being SiOx or SiN due to the overlapping of absorption bands with each other. Identifying these alloys implies that this system can be applied to any class of semiconductors by developing library for them and adding the correct rules for these alloys.

The rapid developments in the instrumentation and computer technology made great advances in information processing and obtaining more meaningful results from the raw data. The automated spectra interpretation can be made by using library search or rule-based (expert) systems. Library search systems do not use explicitly formulated interpretation rules; they rather rely on a large library of reference spectra from compounds of known structure. The unknown sample spectrum is compared to all the library spectra and the hit lists of the references are probably identical or similar to the unknown. Library search is the most useful technique but not always gives correct results if you do not have a comprehensive library. The rule-based system does not need a library. One of the most important parts of any expert system is its knowledge base, which includes information about the problem domain [21]. This information can be stored in a variety of forms, of which simple if-then rules are the most common.

There is a lot of research on the automated spectrum interpretation, and use of the expert systems in this process [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34]. But there is not enough work on the interpretation of amorphous or polycrystalline semiconductors because of the following reasons. The semiconducting a-Si materials are being made at non-stochiometric ratios since the required parameters such as band gap, mobility or conductivity are obtained at non-stochiometric ratios. Moreover, the production techniques for these compounds are diverse, such as glow discharge, sputtering, vacuum evaporation, electron beam evaporation, electrolysis, etc. And IR spectra of the semiconductors depend on the production method [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20]. Due to these reasons it is difficult to form a full library to represent all these silicon alloys which are diverse regarding to the compositional ratios, deposition systems and conditions. To our knowledge, there is no commercially available library and software to identify amorphous semiconductor alloys. The library search system alone cannot give correct answer for the spectra interpretations. For this reason, the library search and rule-based systems need to be used together to develop an hybrid system [27], [33]. Such a system is formed and applied to the a-Si alloys spectra, which were obtained from the previous work [35], [36].

Section snippets

Development of the reference library

There are many infrared spectra for amorphous semiconductors; therefore, it is difficult to make a full reference library. All available spectra in printed form, which were used as unknown spectra, were scanned and converted into a computer readable format using an interfacing program. The details were given in the earlier work [35], [36]. The reference library was developed for a-SiC, a-SiN, a-Si, a-SiOx alloys. For each reference spectra, the peak positions of the main and sub-bands,

Results

The developed system was applied to the IR analysis of a-Si alloys and about 95% of the time the correct hit ratio obtained. The results of the interpretation of unknown spectra are shown in Table 1, Table 2, Table 3, Table 4. In these tables, the first column shows the names of the spectra which are taken as unknowns. The second, third, fourth and fifth columns show SQI values calculated for the reference spectra of a-SiC, a-SiN, a-Si and a-SiOx, respectively. The sixth column shows the

Discussion

It is difficult to develop the computer-aided structure identification system for semiconducting amorphous materials due to the reasons discussed in the introduction section.

In this work, the library search system was used together with some rules and similarity quality index calculated for the certain regions. The candidate compound was selected using minimum SQI value (library searching). If there exists two candidates with similar SQIs then sub-band check (rule-based classifying) was used to

Conclusion

Fourier transform infrared spectroscopy is a viable tool for qualitative and quantitative characterization of amorphous silicon alloys. There is no commercial program which, for amorphous semiconductor materials, can assist the researchers for easier identification. The developed program enables the user to identify the spectra with computer aid.

The developed system can be improved for quantitative analysis by adding enough information to the rule based. For example, it can give information

References (37)

View full text