Aluminum schottky contacts to n-type 4H-SiC

Journal of ELECTRONIC MATERIALS, Vol. 31, No. 10, 2002

Regular Issue Paper

Aluminum Schottky Contacts to n-Type 4H-SiC WILLIAM R. HARRELL,1,2 JINGYAN ZHANG,1 and KELVIN F. POOLE1 1.—Center for Silicon Nanoelectronics, Holcombe Department of Electrical and Computer Engineering, Clemson University, Clemson, SC 29634-0915. 2.—E-mail: [email protected]

Aluminum Schottky contacts to n-type 4H-SiC were fabricated, and the results of electrical measurements on the devices are presented along with the Schottky barrier height and ideality factor. Furthermore, two different samples having different surface treatments were investigated. The barrier heights obtained from current-voltage (I-V) measurements on both samples are essentially identical, with an average value of 0.64 eV. The ideality factor of Sample 2, which had an HF-dip surface treatment, had an average value of 1.66. The ideality factor of Sample 1, which had a surface treatment similar to the other sample but with the addition of an RCA-like clean, had an average value of 2.60. Measurements on some control devices with Au contacts showed that no significant Fermi-level pinning occurred, and the barrier-height results for all devices approached the Schottky–Mott limit. Capacitance-voltage (C-V) measurements on the Al devices from Sample 1 yielded barrier heights that were much smaller than those obtained from the I-V curves; however, for Sample 2, the barrier heights obtained from both C-V and I-V methods were comparable. Based on these results, we propose a model that accounts for the measured results in terms of a thin interfacial layer produced in Sample 1, which screens the depletion region from the applied electric field. This model yields a possible explanation of how the two surface treatments result in devices with the same barrier height but different ideality factors. The results in this paper represent some of the first experimental data on Al/4H-SiC Schottky contacts. Key words: 4H-silicon carbide, Al contacts, Schottky diodes

INTRODUCTION Silicon carbide (SiC) is an emerging, compoundsemiconductor material with many outstanding physical properties that make it an excellent candidate for use in many important electronic applications. The wide bandgap, high-thermal conductivity, high-breakdown field, and robust mechanical properties of SiC make it an attractive material for hightemperature, high-power, and high-frequency electronic devices.1–3 Metal-semiconductor contacts are fundamental building blocks of any integrated circuit technology, so a knowledge and understanding of the processing and subsequent operation of these contacts is very important. Furthermore, the 4H polytype of SiC has generated much interest re(Received January 16, 2002; accepted July 22, 2002) 1090

cently because of its outstanding electronic properties. For example, 4H-SiC has an electron mobility approximately twice that of 6H-SiC, as well as a higher hole mobility and a slightly wider bandgap.4 Aluminum is a technologically important metal in the semiconductor industry, and although several reports have been published on Al contacts to 6H SiC, there are very few results for Al contacts to 4H SiC. In this paper, results are reported on electrical measurements performed on aluminum Schottky contacts to n-type 4H-SiC, and the results obtained from samples with two different surface treatments are compared. In the remainder of this paper, the details of the study on Al/4H-SiC contacts will be presented. A description of the starting materials and fabrication processes will be contained in the next section, with special attention given to the surface preparation

Aluminum Schottky Contacts to n-Type 4H-SiC

sequence. The methods used in performing electrical measurements will also be briefly outlined. In the third section, the experimental results will be presented along with a detailed discussion. In particular, results of current density-voltage (J-V) measurements on the devices and extraction of the barrier height and ideality factor for two different surface treatments will be presented. The implications of the results will be discussed by comparing the results to similar devices fabricated with gold contacts and also by examining the results of capacitancevoltage (C-V) measurements on the devices. Finally, the primary conclusions of this work will be summarized. SAMPLE PREPARATION AND EXPERIMENTAL PROCEDURE The starting materials used in this study consisted of n-type 4H-SiC wafers, Si-face, doped to a concentration of 5.4 ⫻ 1018 cm⫺3. Each sample had an n-type epitaxial layer grown to a thickness of 5 ␮m with a doping concentration of 5 ⫻ 1015 cm⫺3. All wafers, which were diced into quarter-inch squares, were purchased from Cree, Inc.5 In this study, we used two different surface-treatment processes, which will be referred to as Surface Treatment 1 and Surface Treatment 2. Each of these processes are described subsequently. Surface Treatment 1 Surface preparation and cleaning consisted of three sequential steps. First, each sample was degreased in acetone and methanol. Then, the samples were dipped in a solution of diluted HF (48%), consisting of H2O:HF (25:1), for 2 min to remove any native oxide. Finally, each sample was cleaned with agitation in a mixture of deionized (DI) water:H2O2: Baker Clean (5:0.22:1) at a temperature of 60°C for 15 min. The Baker Clean is essentially a modified RCA solution6 and was obtained from J.T. Baker, Inc.7 After each cleaning process, the samples were rinsed in DI water for a total of 6 min. The material receiving this surface treatment will be referred to as Sample 1.

1091

contacts. Subsequently, the samples were degreased in acetone and methanol to remove any contamination occurring from the diffusion pump during the evaporation process. Finally, the as-deposited Ni contacts were annealed in a furnace at 1000°C for 3 min in forming gas (10% H2 with argon balance) to minimize contact resistance and prevent surface oxidation. The shadow mask used for the Ni contacts was designed to produce one large dot and two smaller test dots on the back side of the samples. Current-voltage (I-V) measurements, performed between the two test dots, confirmed that the Ni contacts were ohmic, with a resistance well below 1 ⍀. Schottky contacts were formed on the front-side epitaxial layer of the samples by evaporating approximately 1-␮m-thick by 630-␮m-diameter aluminum dots. The shadow mask used for the Schottky contacts was designed to produce an array of 14 diodes on each sample. Electrical measurements were performed on the SiC samples inside an enclosed probe station manufactured by Micromanipulator Co., Inc. (Carson City, NV). Current-voltage measurements were performed with an HP 4156B Precision Semiconductor Parameter Analyzer. High-frequency (1 MHz) C-V measurements were performed with a DLTS system manufactured by Sula Technology, Inc. (Ashland, OR). RESULTS AND DISCUSSION Current-voltage measurements were performed on 14 diodes from Sample 1 and Sample 2, for a total of 28 diode measurements. The measured J-V curves for Samples 1 and 2 under forward bias are shown in Figs. 1 and 2, respectively. In Figs. 1 and 2, the range of the J-V curves measured on both samples is illustrated by showing the maximum and minimum curves, with all other measurements falling in between the two curves shown. The leveling off of the

Surface Treatment 2 Surface preparation and cleaning consisted of two sequential steps. These two steps were identical to the first two steps of Surface Treatment 1. Thus, the only difference between the two surface-treatment processes is that Surface Treatment 2 did not include the Baker Clean step. The material receiving this surface treatment will be referred to as Sample 2. Metal contacts to the SiC samples were formed by vacuum evaporation at a base pressure of 5 ⫻ 10⫺6 torr, using shadow masks to form circular dots. Ohmic contacts were formed on the back side of the samples with approximately 1700 Å of nickel. The relatively high doping concentration of the substrate was chosen to produce low-resistance ohmic

Fig. 1 The J-V curves of Al/4H-SiC Schottky diodes fabricated on Sample 1. The two curves represent the range of measurements on 14 devices.

1092

Fig. 2 The J-V curves of Al/4H-SiC Schottky diodes fabricated on Sample 2. The two curves represent the range of measurements on 14 devices.

J-V curves in Figs. 1 and 2 is due to the maximum current-compliance limit on the HP 4156B, which is 100 mA. Two observations can readily be made simply by comparing the forward J-V curves from the two samples. First, there is a tighter distribution in the J-V curves from Sample 2 as compared to Sample 1. In addition, the curves from Sample 2 have a higher slope for most of the voltage range observed; and, although the Sample 1 currents are somewhat higher at lower voltages, the Sample 2 currents tend to be higher for most of the voltage range. Thus, it is apparent that the surface treatments have a significant impact on the behavior of these diodes. Some explanations for these observations will be discussed subsequently in this paper. To analyze the measured data obtained from these Al/4H-SiC Schottky diodes, the Schottky barrier height, ␾B, and the ideality factor, n, must be determined. For an ideal Schottky barrier diode, transport over the barrier is governed by thermionic emission,8 and the current density, J, can be described by the Richardson–Dushman equation:9 q␾B qV J ⫽ A*T2 exp a⫺ b exp c d (1) kT nkT where A* is the effective Richardson’s constant, T is the absolute temperature, q␾B is the Schottky barrier height in eV, k is Boltzmann’s constant, n is the ideality factor, and V is the applied voltage. Here, we have neglected the 1 in the argument of the second exponential because it is negligible for forward voltages greater than about 3 kT/q. The effective Richardson’s constant for 4H-SiC was taken to be A* ⫽ 146 A/(K2cm2).10,11 Thermionic-emission theory predicts that a semilog J-V plot will be linear for forward bias voltages. Observation of Figs. 1 and 2 indicates that only part of the measured J-V curves are linear, and thus, more than one mechanism must be in operation

Harrell, Zhang, and Poole

depending on the voltage range. We have described these mechanisms elsewhere by analyzing the entire J-V curve and determined that, in addition to thermionic emission at lower voltages, space-chargelimited emission dominates at higher voltages until series resistance takes over at approximately 1 V.12 In this paper, however, we are only concerned with extracting the Schottky parameters from the lower voltage data to characterize the Al contacts. Equation 1 can also be written as qV (2) J ⫽ Js exp c d nkT where Js ⫽ A*T2 exp (⫺q␾B/kT) is the reverse-saturation current density. Using a standard procedure, Eq. 2 was employed to model the measured J-V data; the reverse-saturation current density, Js, was obtained from the J intercept of the linear portion of a semilog J-V plot; and the barrier height was calculated from that result. The ideality factor, n, was then determined from a least-squares fit to Eq. 2. The J-V curves from Sample 2 (Fig. 2) have a noticeable linear region from approximately 0.1–0.3 V, and the correlation coefficient for the best fit line to the data in this voltage range was 0.997–0.999 for the 14 devices measured. Typical values of Js extrapolated from these curves were on the order of 10⫺4 A/cm2 or less, corresponding to a reverse bias of 20 V. However, from inspection of the J-V curves from Sample 1 (Fig. 1), a linear region is not obvious. Therefore, we used the same voltage range as was used for Fig. 2 to extrapolate the barrier height and ideality factor to compare the two samples consistently. The results for both samples of Al devices are shown in Table I. The results shown in Figs. 1 and 2 Table I. Barrier Heights and Ideality Factors Obtained from J-V Measurements on Al/4H-SiC Schottky Contacts Sample 1

Average Maximum Minimum Standard deviation

Sample 2

␾B (eV)

n

␾B (eV)

n

0.64 0.63 0.63 0.62 0.61 0.60 0.59 0.58 0.58 0.59 0.60 0.58 0.59 0.59

2.52 2.51 2.51 2.60 2.57 2.60 2.62 2.66 2.68 2.67 2.61 2.69 2.63 2.61

0.67 0.67 0.67 0.66 0.69 0.69 0.69 0.66 0.67 0.67 0.66 0.66 0.67 0.66

1.66 1.64 1.69 1.77 1.62 1.57 1.58 1.66 1.65 1.67 1.73 1.69 1.67 1.67

0.60 0.64 0.58 0.02

2.60 2.69 2.51 0.06

0.67 0.69 0.66 0.01

1.66 1.77 1.57 0.05

Aluminum Schottky Contacts to n-Type 4H-SiC

and in Table I represent a larger body of our work on Al/4H-SiC devices and are some of the first data published on such devices. Although there have been a few published reports on Al contacts to 4HSiC,13,14 in the former report, p-type SiC was used along with a thin surface implant, and in the latter report, large area contacts were diffusion welded, while no analysis of the Schottky parameters was performed. Several immediate observations can be made from the data in Table I. First, within each sample of devices, there is very little variation in the measured barrier heights and ideality factors, indicating the uniformity of the devices undergoing the same process sequence. With a variation in the measured barrier heights of a few hundredths of an eV and a variation in the ideality factors of 0.1 to 0.2, and considering the standard deviations shown in Table I, the consistency of the data is good. The barrierheight measurements are accurate to the nearest 0.1 eV, while the ideality factors are accurate to within 0.1. In addition, the results were repeatable when making multiple measurements on the same samples over a period of several months, with no significant differences from the results shown in Table I. Another noteworthy observation is that there is no significant difference in the barrier heights between the two samples, although there is slightly more spread in the ␾B data for Sample 1, as seen from the calculated standard deviation. Thus, the barrier height of all 28 devices has an average value of approximately 0.64 eV. This similarity in the data indicates that the different surface treatments had little effect on the Schottky barrier height of these devices. However, this conclusion does not apply to the ideality-factor data. Although there is little variation in the measured values for n within each sample, there is a noticeable difference between the two samples. For Sample 1, the average value is n ⫽ 2.60, and for Sample 2, the average value is n ⫽ 1.66, which yields a difference of 0.94 in the ideality factors of the two samples. This data indicates that the different surface treatments do have an effect on the ideality factor and, thus, the behavior of the devices. Furthermore, this difference is consistent with the J-V curves of Figs. 1 and 2, as stated previously. Thus, the different surface treatments affect the ideality factor but not the barrier heights of these devices. To attempt to explain this result will require further analysis and modeling, which will be presented subsequently in this paper. Although there are very few results on Al/4H-SiC contacts in the literature to compare these results to, it is insightful to present a brief comparison to Al/6H-SiC results because 6H- and 4H-SiC are similar in many ways. For one thing, the two polytypes both have a hexagonal structure. More importantly for metal/SiC structures, they have similar electron affinities, with q␹ ⫽ 3.3 eV for 6H-SiC2,15 and q␹ ⫽ 3.6 eV for 4H-SiC.16 Therefore, at least within the Schottky–Mott limit, one might anticipate compara-

1093

ble barrier heights for contacts to 4H- and 6H-SiC. Several authors have reported barrier heights for Al/6H-SiC contacts that compare well with the average 4H result of q␾B ⫽ 0.64 eV presented here. Some examples are q␾B ⫽ 0.678 eV,17 0.54 eV,18 and 0.61 eV.19 In particular, the Schottky contacts reported in Ref. 17 were very similar to our Sample 2, with a background doping on the order of 1018 cm⫺3, a similar HF surface treatment, thermal evaporation of aluminum, and barrier-height extraction from I-V measurements. Their result of q␾B ⫽ 0.678 eV compares well with the average value of q␾B ⫽ 0.67 eV from our Sample 2, shown in Table I. On the other hand, there is some spread in the reported barrierheight results for Al/6H-SiC contacts, with a lower range of 0.2–0.3 eV20 and an upper range of 1.45–1.7 eV.21,22 When examining ideality factors, one finds a much wider variation in the literature. Reported values of n range from slightly above unity to the relatively high values of 5.617 to 13.3.18 This spread is significant and most likely due, in part, to variations in surface treatment and processing conditions. However, values for n shown in Table I, which range from n ⫽ 1.57 to n ⫽ 2.69, do compare well with some reports from the literature on ideality factors for Al/6H-SiC contacts, such as n ⫽ 2–3,19 1.72,20 1.5–2,21 2–2.3,22 and 1.68.23 Although there is a range of Schottky parameters for Al/6H-SiC contacts reported in the literature, many of these results fall within the range of results presented in this paper, and this provides at least some basis for comparison in the absence of any Al/4H-SiC results. More detailed comparisons between Schottky contacts to 6H- and 4H-SiC would be beneficial and interesting to the research community, and we are currently pursuing such investigations. To further investigate these results on Al/4H-SiC diodes, several gold contacts were fabricated on both SiC samples as controls. All experimental conditions were identical to facilitate comparison between the Al/ and Au/4H-SiC contacts. Using the same I-V measurement and analysis methods as for the Al devices, the barrier height and ideality factor were determined for each Au device. There was essentially no variation observed in ␾B or n between the two samples. For the Au/4H-SiC diodes, the average value of the barrier height was 1.07 eV, while the average value of the ideality factor was 1.23, both of which are noticeably different from the Al/4H-SiC results. Of particular interest is the fact that the barrier height for the Au devices is larger than that for the Al devices. In the Schottky–Mott limit for metal/semiconductor junctions, valid in the absence of a significant density of interface states, the barrier height in electron-volts is given by q␾B ⫽ q⌽M ⫺ q␹

(3)

where ⌽M is the metal work function. Although Eq. 3 is for an ideal Schottky contact, it is useful for examining measured barrier heights. Work functions for many metals are readily obtained from many ref-

1094

Harrell, Zhang, and Poole

erences, but often these measurements are for nearideal cases. The metallization process in this study was implemented with an oil-based vacuum-evaporation system, which produces less than ideal conditions and metal/semiconductor junctions; therefore, an appropriate metal work function should be used. A table of values for metal work functions, averaged from many experimental results on samples that were prepared in a range of vacuum systems from poor to excellent, can be found in Ref. 24. The work functions for Al and Au are 4.20 eV and 4.89 eV, respectively. Using these values for q⌽M along with q␹ ⫽ 3.6 eV in Eq. 3, the theoretical barrier heights are q␾B[Al/4H-SiC] ⫽ 0.60 eV q␾B[Au/4H-SiC] ⫽ 1.29 eV

(4)

For the Al contacts, the theoretical result in Eq. 4 is very close to the average experimental result of q␾B ⫽ 0.67 eV, differing by only 0.07 eV. For the Au contacts, the theoretical result in Eq. 4 is near the average experimental result of q␾B ⫽ 1.07 eV, differing by 0.22 eV. The difference between the theoretical and experimental values of q␾B for the Au contacts could be partially due to image-force barrier lowering, but this effect is normally not this large. The primary point to be made with these results is the trend in the measured barrier heights. An increase in metal work function yields an increase in the measured barrier height because the barrier height of the Au contacts is greater than that of the Al contacts. Therefore, no significant Fermi-level pinning by interface states is observed in these devices, similar to the results of Waldrop for contacts to 6H-SiC.20 One observation mentioned previously that needs further analysis is the fact that the barrier heights obtained for the two Al/4H-SiC samples are essentially identical, while the ideality factors are noticeably different. The two surface treatments obviously affect the devices differently, as seen from the variations in the measured J-V curves, shown in Figs. 1 and 2, as well as in n; however, this effect evidently does not change the Schottky barrier height. This observation can be explained by considering the results of high frequency C-V measurements on the devices. The C-V measurements were performed at 1 MHz on each device in reverse bias, and the barrier heights were extracted from the intercept of the 1/C2-V curve. The barrier heights obtained from the C-V measurements were consistent within each sample, with total variations of 0.1 eV for Sample 1 and 0.02eV for Sample 2. The average value for Sample 1 was q␾B ⫽ 0.39 eV, while the average value for Sample 2 was q␾B ⫽ 0.65 eV. For Sample 2, the barrier heights determined from the C-V measurements are very close to the results obtained from the J-V measurements. Recall that Sample 2 did not have the Baker Clean step performed during the surface-treatment process. For Sample 1, however, the barrier heights determined from the C-V meas-

urements are much less than the results obtained from the J-V measurements. Schottky barrier heights are often overestimated from C-V measurements if there is a thin oxide layer at the interface between the metal and semiconductor; thus, sometimes both I-V and C-V measurements are performed, and the “real” barrier height is taken to lie somewhere between the two results.10 The reason for this is clearly seen from the basic equation for capacitance at a metal/semiconductor junction. The square of the inverse of junction capacitance for a Schottky contact is given by25 1 2 2(Vbi ⫹ VR) b ⫽ (5) C¿ qesNd where C⬘ is the junction capacitance per unit area, Vbi is the built-in voltage, VR is the magnitude of the reverse voltage, ␧s is the dielectric constant of the semiconductor, and Nd is the donor dopant concentration. An oxide layer at the interface would introduce an extra capacitance, Cox, in series with the depletion capacitance, Cdep. This would tend to reduce the total measured capacitance, C⬘, which would imply an increase in Vbi and, thus, an increase in the calculated barrier height. However, the difference in ␾B obtained from the J-V and C-V measurements for Sample 1 in this study is in the opposite direction. A plausible explanation for this observation is that the Baker Clean step used for Sample 1 results in a different surface termination than that of Sample 2, where the HF dip was the last chemical on the surface. Thus, the devices fabricated on Sample 1 evidently have a very thin, interfacial layer that affects both n and C⬘. The added resistance, caused by this interfacial layer, for carrier transport across the junction could explain the significantly higher ideality factor for Sample 1. Furthermore, if the interfacial layer screens the depletion region in the SiC from the field, then the depletion width would be smaller in Sample 1 than in Sample 2 for the same applied reverse bias. Because the depletion capacitance is inversely proportional to the depletion width, Cdep would increase. The ultimate effect of the screening layer then would be to reduce 1/C2, which, according to Eq. 5, would tend to reduce Vbi and, therefore, reduce the calculated barrier height. Such an interfacial layer evidently has little or no effect on the actual barrier height, as determined from the J-V measurements. Finally, the C-V measurements were performed on the Au/4H-SiC diodes, and there was essentially no difference in measured barrier height between the two samples. The average value obtained of q␾B ⫽ 1.15 eV was in agreement with the I-V results for both samples. Because gold is normally not very reactive, which is why it is often used as a control metal, there is less probability of forming interfacial layers between the Au and the surface it is deposited upon. Aluminum is a much more reactive and less stable metal, and from these electrical results, it appears that the Al reacts with the surface of Sample 1, resulting in some type a

Aluminum Schottky Contacts to n-Type 4H-SiC

of interfacial layer. This model provides a reasonable explanation for the observed measurements; however, additional, more detailed measurements and modeling may provide deeper insights, and further studies are currently in progress. CONCLUSIONS In this study, Al/4H-SiC Schottky contacts were fabricated, and the electrical characteristics and Schottky parameters were presented. This work represents some of the first data published on Al/4H-SiC devices. There is some general consistency between these results and previously published results on Al/6H-SiC contacts, even though there is some significant spread in the published data. Measurements on Au/4H-SiC contacts show that no significant Fermi-level pinning occurs in these devices, and the barrier heights for both Au and Al contacts are close to the Schottky–Mott limit. In addition, it is apparent that the different surface treatments had a significant effect on the behavior of the Al Schottky contacts, as observed in the J-V characteristics and the extracted Schottky parameters. Sample 1, which received the Baker Clean, had an average ideality factor of 2.60, while Sample 2, which only had an HF dip, had an average ideality factor of 1.66. However, although the ideality factors for the devices on Sample 1 were significantly larger than for devices on Sample 2, the barrier heights were approximately the same for all devices. An explanation for this observation, based on C-V measurements, was attributed to different surface terminations for the two samples. The depletion region was evidently screened from the applied electric field by an interfacial layer in Sample 1. This work represents some new experimental results on Al/4HSiC Schottky contacts. More detailed measurements and modeling studies are currently under way, including investigations of the effects on the interface of ultraviolet-assisted, metal-organic chemical-vapor deposition of Al onto 4H-SiC.

1095

REFERENCES 1. J.B. Casady and R.W. Johnson, Solid-State Electron. 39, 1409 (1996). 2. K. Järrendahl and R.F. Davis, Semiconductors and Semimetals, ed. Y.S. Park (New York: Academic Press, 1998), vol. 52, pp. 1–18. 3. C.E. Weitzel, J.W. Palmour, C.H. Carter, K. Moore, K.J. Nordquist, S. Allen, C. Thero, and M. Bhatnagar, IEEE Trans. Electron. Dev. 43, 1732 (1996). 4. P.G. Neudeck, J. Electron. Mater. 24, 283 (1995). 5. Cree, Inc., Durham, NC. 6. W.A. Cady and M. Varadarajan, J. Electrochem. Soc. 143, 2064 (1996). 7. J.T. BAKER, Inc., Philipsburg, NJ. 8. E.H. Rhoderick and R.H. Williams, Metal-Semiconductor Contacts, 2nd ed. (Oxford, United Kingdom: Clarendon, 1988), pp. 89–124. 9. S.M. Sze, Physics of Semiconductor Devices, 2nd ed. (New York: Wiley, 1981), pp. 254–263. 10. A. Itoh and H. Matsunami, Crit. Rev. Solid State Mater. Sci. 22, 111 (1997). 11. S.K. Lee, C.M. Zetterling, and M. Östling, J. Appl. Phys. 87, 8039 (2000). 12. J. Zhang, W.R. Harrell, R. Singh, and K.F. Poole, submitted for publication. 13. V. Khemka, K. Chatty, T.P. Chow, and R.J. Gutmann, Mater. Sci. Forum 338–342, 1211 (2000). 14. O. Korolkov and T. Rang, Mater. Sci. Forum 353–356, 603 (2001). 15. L.M. Porter and R.F. Davis, Mater. Sci. Eng. B34, 83 (1995). 16. J. Campi, Y. Shi, Y. Luo, F. Yan, and J.H. Zhao, IEEE Trans. Electron. Dev. 46, 511 (1999). 17. Y.G. Zhang, X.L. Li, A.Z. Li, and A.G. Milnes, Inst. Phys. Conf. Ser. 142, 665 (1996). 18. C.V. Reddy, S. Fung, C.D. Beling, and G. Brauer, Inst. Phys. Conf. Ser. 142, 669 (1996). 19. K. Matsumoto, Y. Chen, J. Kuzmik, and S. Nishino, Mater. Sci. Forum 264–268, 925 (1998). 20. J.R. Waldrop, R.W. Grant, Y.C. Wang, and R.F. Davis, J. Appl. Phys. 72, 4757 (1992). 21. S.H. Hagen, J. Appl. Phys. 39, 1458 (1968). 22. K. Yasuda, T. Hayakawa, and M. Saji, IEEE Trans. Electron Dev. 34, 2002 (1987). 23. A.M. Strel’chuk and M.G. Rastegaeva, Mater. Sci. Eng. B46, 379 (1997). 24. S.P. Parker, McGraw-Hill Encyclopedia of Physics, 2nd ed. (New York: McGraw-Hill, 1993), p. 1545. 25. D.A. Neamen, Semiconductor Physics and Devices: Basic Principles, 2nd ed. (Chicago, IL: Irwin, 1997), p. 313.