Charge Density in Atmospheric Pressure Chemical Vapor Deposition TiO 2 on SiO 2 -Passivated Silicon Keith R. McIntosh, a,z Simeon C. Baker-Finch, a Nicholas E. Grant, a Andrew F. Thomson, a Sonita Singh, a and Iain D. Baikie b a Center for Sustainable Energy Systems, Australian National University, Canberra ACT 2600, Australia b KP Technology Limited, Wick KW1 5EH, United Kingdom The charge density of a TiO 2 film deposited on a SiO 2 -passivated silicon wafer is determined. The TiO 2 is deposited by atmo- spheric pressure chemical vapor deposition at 400°C, and the SiO 2 is grown thermally at 950°C. This TiO 2 –SiO 2 stack is a useful coating for the front surface of a silicon solar cell, as it has a high optical transmission and a low density of interface states D it E at the SiO 2 –Si interface. While these properties are beneficial to high efficiency solar cells, so too is a large charge density, as what occurs in Si 3 N 4 –SiO 2 +10 12 cm -2 and Al 2 O 3 –SiO 2 -10 13 cm -2 stacks. The D it Eand charge density of TiO 2 -coated and SiO 2 -passivated silicon are evaluated by capacitance–voltage and Kelvin probe measurements. The charge density of the TiO 2 is within the conservative limits of 8.5 and -1 10 11 cm -2 after deposition and of 10 and +1 10 11 cm -2 after a subsequent 800°C oxygen anneal. Photoconductance measurements suggest that the dangling-bond defects at the SiO 2 –Si interface are predominantly donorlike and, hence, that the change density in the TiO 2 is closer to the upper limits less negative; this charge is too small to benefit solar cells. © 2009 The Electrochemical Society. DOI: 10.1149/1.3216029All rights reserved. Manuscript submitted January 16, 2009; revised manuscript received August 4, 2009. Published September 14, 2009. The recombination rate at the interface between a semiconductor and an insulator U s depends strongly on the insulator’s charge den- sity Q i . Figure 1 plots U s Q i for a SiO 2 –Si interface over a range of midgap interface-state densities D it0 when all dangling-bond states are either adonorlike or bacceptor-like. The figure shows that in general, U s decreases with increasing Q i , U s decreases with de- creasing D it0 , and U s depends more strongly on Q i when D it0 is small. The first trend results from the suppression of recombination via the limitation of either the hole for Q i 0or the electron for Q i 0concentration; the second results from the suppression of recombination via the reduction in recombination centers; and the third results from less charge associated with the interface states Q it . Exceptions to these trends arise when the interface-state density is very large e.g., D it0 =3 10 12 cm -2 eV -1 making the associated interface charge Q it comparable to Q i . Similar plots are described in more detail by Aberle et al. 1 The calculations and the variables associated with Fig. 1 are described in the Appendixes. The dependence of U s on D it0 and Q i is important to the manu- facture of high efficiency photovoltaic solar cells. Their performance depends greatly on the recombination at the illuminated surface, near which most electron–hole pairs are generated. In all high effi- ciency designs, 2-6 a hydrogenated thermal SiO 2 is used as a passi- vation layer nearest the silicon, providing a small D it0 of 10 10 cm -2 eV -1 and a small Q i of 5–20 10 10 cm -2 . 7 As shown in Fig. 1, this Q i is insufficient to reduce U s substantially, and so a surface diffusion is also used to reduce the minority carrier concen- tration. In most cells, a surface diffusion is required to form the p-n junction, but in either case, it helps to lower U s as Q i is insufficient. However, with sufficient Q i a smaller U s can be attained without a surface diffusion 8 presumably because the diffusion introduces ad- ditional interface states leading to a higher D it0 . In addition to the SiO 2 layer, high efficiency cells incorporate one or two more dielectrics for their antireflection properties. Such dielectrics include titanium dioxide TiO 2 , 2,9-11 aluminum oxide Al 2 O 3 , 11,12 silicon nitride Si 3 N 4 , 6 amorphous silicon nitride SiN x , 7,13 magnesium fluoride MgF 2 , 3 and zinc sulfide ZnS. 3,9 Although not always considered, the charge density of these dielec- trics also contributes to Q i and therefore influences U s and the cell efficiency. The charge density is large and positive in SiN x and Si 3 N 4 10 12 cm -2 7,13 and large and negative in Al 2 O 3 –SiO 2 stacks 10 13 cm -2 , 9 but the charge density of either TiO 2 or TiO 2 –SiO 2 is not accurately ascertained despite the application of TiO 2 as an antireflection coating on solar cells since the 1970s 10 albeit in- creasingly rarely. To date, quantification of TiO 2 ’s charge density has been made by capacitance–voltage CVmeasurements, 14-16 a technique that determines the sum of Q i and Q it at flatband 17 but not an indepen- dent value of Q i . Because the Q it of such samples is not necessarily negligible and the insulating surface of a solar cell does not operate at a flatband potential, these studies do not provide a useful assess- ment of Q for photovoltaic applications. This paper concerns the measurement of Q i in a TiO 2 –SiO 2 –Si structure, where TiO 2 has been deposited by atmospheric pressure chemical vapor deposition APCVD. A combination of CV and Kelvin probe KPtechniques was applied to extract Q i , an approach that is applicable to a semiconductor–insulator structure with a small Q i . The results were supported by photoconductance PCmeasure- ments. Sample Preparation Figure 2 presents diagrams of the samples used in this work. All samples were fabricated from 100float-zoned n-type silicon wa- fers with nominal resistivities of 4.6–5.4 cm. Before thermal oxidation, the samples received a 20 min etch in tetramethylammo- nium hydroxide at 75–85°C to remove saw damage, a 3 min hydrofluoric acid:HNO 3 chemical polish at room temperature, and z E-mail: KRMcIntosh@gmail.com (b) (a) Figure 1. The solid lines plot the surface recombination rate U s as a function of insulator charge density Q i and midgap interface-state density D it0 when all dangling-bond states are either adonorlike or bacceptor-like. The calculations were made for an n-type Si semiconductor with a donor concen- tration of N D = 10 15 cm -3 and an excess carrier density in the bulk of n b = 10 15 cm -3 ; the remaining parameters are listed in Appendix B. The dashed line plots U s of the experimental samples determined by PC decay, as described in the PC Results section. Journal of The Electrochemical Society, 156 11G190-G195 2009 0013-4651/2009/15611/G190/6/$25.00 © The Electrochemical Society G190 Downloaded 14 Sep 2009 to 150.203.43.180. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp