Solid State Phenomena Vols. 18-19 (25) pp 621-626 Online available since 25/Dec/15 at www.scientific.net (25) Trans Tech Publications, Switzerland doi:1.428/www.scientific.net/ssp.18-19.621 Infrared Absorption Measurement of Carbon Concentration Down to 1x1 14 /cm 3 In CZ Silicon N. Inoue 1, a and M. Nakatsu 1, b 1 Research Institute for Advanced Science and Technology, Osaka Prefecture University 1-2, Gakuencho, Sakai, Osaka, 599-8, Japan a inouen@riast.osakafu-u.ac.jp, b nakatsu@riast.osakafu-u.ac.jp Key Words: Silicon, Carbon, Infrared Absorption Abstract. Measurement of carbon concentration in CZ silicon by infrared absorption spectroscopy was examined. Noise level was suppressed down to 1-4 in unit of. Residual differential absorption between the sample and reference was removed by fitting the phonon absorption spectrum to the background absorption spectrum. The effect of narrowing of absorption spectral range was examined. As a result, it was possible to measure the differential carbon concentration down to about 1 1 14 /cm 3. Measurement of commercial wafer was also established. Introduction Carbon concentration in CZ silicon substrate has decreased below the assumed detection limit of 2x1 15 /cm 3 of infrared absorption spectroscopy (IR) with procedure given in the ASTM standard [1]. We have developed both experimental and arithmetic procedures to measure nitrogen concentration below 1 14 /cm 3 [2]. In this paper we applied these techniques to carbon measurement and improved these methods and successfully measured carbon concentration ([C]) down to about 1 14 /cm 3. Measurement on commercial wafers was also examined. Experimental and Results Two millimeter thick, double side mirror polished samples with various carbon concentrations were prepared. Lowest concentration sample was used as a reference for difference method. A Fourier transform IR spectrometer with a TGS detector was used at room temperature. Wavenumber resolution was 2cm -1. Sample chamber was filled with dried N 2 gas flow. The necessary signal to noise ratio was estimated as follows. In the case that carbon concentration is 1 1 15 /cm 3, carbon peak is (1 1 15 /8 1 16 [1]/2.33/5=). It is necessary to make S/N better than the carbon peak. The measurement condition of the sample was made as close as possible to that of the reference sample and the number of repetition of the scan was increased. Figure 1 shows an example of differential spectrum between the two measurements on the same sample. We were able to suppress the noise down to about.1. All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of TTP, www.ttp.net. (ID: 13.23.136.75, Pennsylvania State University, University Park, United States of America-3/6/14,1:24:49)
622 Gettering and Defect Engineering in Semiconductor Technology XI In case of [C] around 1 1 15 /cm 3 or the previous detection limit, there is another origin of error. Carbon peak at 65 cm -1 is located on the shoulder of the huge double phonon peak at 61 cm -1. In the case of high concentration, it is possible to draw a long straight baseline for example from to 64 cm-1 as given previously in the standard [1], because the effect of double phonon peak is small. But in case of low concentration, differential phonon is bigger than the carbon absorption. Therefore it is impossible to draw a straight baseline. The influence of the phonon peak was suppressed by the arithmetic procedures as follows. First, background absorption was made nearly flat by weighted subtraction of phonon absorption with adequate weight. Remaining background differential absorption showed the wave number dependency of that of the phonon absorption. So, next the fraction of phonon absorption was fitted to the background absorption as shown by the thin line in Fig. 2(a). As a result, it was possible to detect the peak of carbon below the traditional detection limit. In case of low carbon concentration, spectral range of absorption peak became narrower than that of the high concentration sample as an example shown by arrows in Figs. 2(a). The end points of the baseline, was for example from 59 to 61 cm -1, as shown by the dotted line in Fig. 2(b). It might neglect the absorption at the foot and results in the underestimation of the carbon concentration. This effect was examined in detail and the correction procedure was developed. It is better to correct for such reduction in case of [C] around 1 1 15 /cm 3. For example spectrum within 59-61 cm-1 underestimates the absorption by about 3% as shown in Fig. 2(b). In case of [C] below 1 15 /cm 3, the background spectrum became irregular. Figure 3 shows an example of spectrum in case of [C] of about 5 1 14 /cm 3. It is difficult to fit the phonon spectrum to the background absorption over a wide spectral range. Therefore fitting was done within the restricted spectral range as shown in Fig. 3. Moreover, carbon absorption peak is usually interfered by steep artificial peaks at about 592 and 612 cm-1 on both sides as marked by arrows in Fig. 3. This is probably related to the change of sign of derivative of the double phonon absorption. It is necessary to ignore these peaks. As the result, we detected of about.5 of the carbon peak. In addition, we investigated the measurement in the case that the differential [C] is about 1 1 14 /cm 3. In this case, the is about.1, nearly equal to the noise level shown in Fig. 1. Fig. 4 shows an example of difference spectrum in the case that the difference concentration is about 2 1 14 /cm 3. We could find carbon absorption peak. It is possible to measure the concentration difference down to about 1 1 14 /cm 3.Finally, we tried to measure the carbon concentration in the commercial wafers. Figure 5 shows the result. We used 4cm-1 of wavenumber resolution to avoid the interference fringe. As the wafer thickness was about.7mm, the was one third of that in the sample with thickness of 2mm. As the sensitivity was about.1 in in our case, we could detect carbon down to 3 1 14 /cm 3 in wafers.
Solid State Phenomena Vols. 18-19 623 Summary In summery, we have examined both experimental and arithmetic procedures to measure carbon concentration in silicon wafers. Signal to noise ratio was improved to.1 in unit of. When the background absorption was wave number dependent, we decided the background absorption at the carbon peak by fitting the reduced phonon absorption to the background absorption spectrum. As a result, it was possible to measure the differential carbon concentration down to about 1 1 14 /cm 3. And we showed that it is possible to measure the carbon concentration in a commercial wafer down to about 3 1 14 /cm 3..8.6.4.2 -.2 -.4 -.6 -.8-64 difference 63 62 61 6 59 w avenum ber ( cm -1 ) Fig. 1 An example of differential spectrum between the repeated measurements on the same sample.
624 Gettering and Defect Engineering in Semiconductor Technology XI.8.7.6 6 4 2.5.4.3.2.8.6.4.2-64 63 62 61 6 59 wavenum ber 64 63 62 61 6 59 w avenum ber Fig. 2(a) An example of differential spectrum from the sample with [C] about 1 1 15 /cm 3. Smooth spectrum was obtained. Wavenumber dependent background absorption was fitted by the reduced phonon absorption shown by the dotted line. (b) A representative carbon absorption spectrum and the straight baseline (solid line) in case of high concentration sample. The dotted line shows the baseline corresponding to a narrower spectral range of Fig. (b), showing the underestimate of..5.45.4.35.3.25.2 5 64 635 63 625 62 615 61 65 6 595 59 585 575 w avenum ber Fig. 3 An example of the differential absorption spectrum from the sample with [C] of 5 1 14 /cm 3. The background absorption became irregular and the fitting was done within the narrow range. Artificial peaks are marked by the arrows.
Solid State Phenomena Vols. 18-19 625.5.45.4.35.3.25.2 64 635 63 625 62 615 61 65 6 595 w avenum ber ( cm -1 ) 59 585 575 Fig. 4 Differential spectrum in case of [C] of 2 1 14 /cm 3..3.2 - -.2 -.3 64 635 63 625 62 615 61 65 6 595 59 w ave n u m be r ( c m - 1 ) 585 575 Fig. 5 An example of absorption spectrum from the sample with carbon concentration about 3 1 14 /cm 3. References [1] Standard test method for substitutional atomic carbon content of silicon by infrared absorption, JEITA EM-353 (JEITA 22), also in ASTM F 1391 and DIN5438/2. [2] M. Nakatsu and N. Inoue, Proc. High Purity Silicon (Electrochemical Society, 24).
Gettering and Defect Engineering in Semiconductor Technology XI 1.428/www.scientific.net/SSP.18-19 Infrared Absorption Measurement of Carbon Concentration down to 1x1 14 /cm 3 in CZ Silicon 1.428/www.scientific.net/SSP.18-19.621