IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 18, NO. 3, MAY 2003

907

An Assessment of Wide Bandgap Semiconductors for Power Devices Jerry L. Hudgins, Senior Member, IEEE, Grigory S. Simin, Member, IEEE, Enrico Santi, Senior Member, IEEE, and M. Asif Khan, Member, IEEE

Abstract—An advantage for some wide bandgap materials, that is often overlooked, is that the thermal coefficient of expansion (CTE) is better matched to the ceramics in use for packaging technology. It is shown that the optimal choice for uni-polar devices is clearly GaN. It is further shown that the future optimal choice for bipolar devices is C (diamond) owing to the large bandgap, high thermal conductivity, and large electron and hole mobilities. A new expression relating the critical electric field for breakdown in abrupt junctions to the material bandgap energy is derived and is further used to derive new expressions for specific on-resistance in power semiconductor devices. These new expressions are compared to the previous literature and the efficacy of specific power devices, such as heterojunction MOSFETs, using GaN are discussed. Index Terms—Breakdown voltage, critical electric field, diamond, GaN, power electronics, SiC, specific on-resistance, wide bandgap.

NOMENCLATURE Pre-factor fit parameter (cm ). Pre-factor fit parameter (V/cm eV or V/cm eV Ionization rate (cm ). Exponential fit parameter (V/cm). Bandgap energy (eV). Electric field as a function of position (V/cm). Critical electric field (V/cm). Permittivity (F/cm). Relative permittivity. Exponent fit parameter. Electron mobility (cm /Vs). Impurity density concentration (cm ). Exponent fit parameter. Specific on-resistance ( cm ). Breakdown voltage (V).

).

I. INTRODUCTION

P

OWER semiconductor devices made from materials with bandgap energies larger than in Si have been touted for many decades. The potential advantages of these wide bandgap devices include higher achievable junction temperatures and

Manuscript received December 13, 2002; revised February 1, 2003. This work was supported by the U.S. Office of Naval Research under Grant N00014-00-1-0131. Recommended by Associate Editor Y. C. Liang. The authors are with the Department of Electrical Engineering, University of South Carolina, Columbia, SC 29208 USA (e-mail: [email protected]). Digital Object Identifier 10.1109/TPEL.2003.810840

thinner drift regions (because of the associated higher critical electric field values) that can result in much lower on-resistance than is possible in Si [1]–[3]. There are however, several disadvantages associated with the use of devices fabricated from wide bandgap materials. Among these is that the ratio of the electron to hole mobility values range much higher than in Si, so that the use of wide bandgap semiconductors for bipolar devices is not desirable. An advantage to the use of some wide bandgap materials that is often overlooked is that the thermal coefficient of expansion (CTE) is better suited to the ceramics used today in packaging technology. It is expected that semiconductor materials that have a closer CTE match to available electrically insulating (thermally conducting) ceramics can more easily be adapted for higher power and wider temperature excursion applications than for Si. A review of the CTE and thermal conductivity of important package materials and selected semiconductors are discussed to highlight the possible advantages of non-Si options. From these thermo-mechanical arguments, the choices of desirable semiconductors are narrowed to SiC, GaN, and C (diamond). , to compare theoretThe use of specific on-resistance, ical limits of different semiconductor materials has been widely accepted as an appropriate figure-of-merit when discussing different semiconductor materials for possible use in producing power devices [1], [2]. The specific on-resistance expression is derived using the theoretical maximum breakdown voltage achievable for an abrupt -junction. The breakdown voltage expression is in turn derived from the relationship for the critical electric field necessary to produce carrier multiplication (avalanche) at the junction. The critical electric field is often expressed in terms of the bandgap energy of the semiconductor. Therefore, it is possible to express the specific on-resistance in terms of the bandgap energy as a means to provide a direct comparison between different semiconductors. It is shown that the commonly used expression relating the critical electric field and the bandgap energy is limited and generally incorrect based on currently available data. It follows that the previously derived expressions for breakdown voltage and specific on-resistance are also limited and generally incorrect. A new expression relating the critical electric field for breakdown in abrupt junctions to the material bandgap energy is derived and is further used to derive new expressions for specific on-resistance in power semiconductor devices. These new relationships are developed based on the available data for a much larger number of different semiconductor materials (critical field values) than was used (or available) to derive the older expressions. Comparisons

0885-8993/03$17.00 © 2003 IEEE

908

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 18, NO. 3, MAY 2003

TABLE I PARAMETERS OF VARIOUS SEMICONDUCTORS AT 300 K

[A] L. I. Berger, Semiconductor Materials, New York: CRC Press, 1997, pp. 105–181. [B] M. Levinshtein, S. Rumyantsev, and M. Shurs, Eds., Semiconductor Parameters, vol. 1, Singapore: World Scientific, 1996. [C] C. M. Wolfe, N. Holonyak, and G. E. Stillman, Physical Properties of Semiconductors, Englewood Cliffs, NJ: Prentice-Hall, 1989, p. 340. [D] R. C. Marshall, J. W. Faust, and C. E. Ryan, Eds., Silicon Carbide—1973, Columbia, SC: Univ. of South Carolina Press, 1974. [E] B. J. Baliga, “Power semiconductor device figure of merit for high-frequency applications,” IEEE Electron Device Lett., vol. 10, no. 10, pp. 455–457, Oct. 1989. [F] V. A. Dmitriev, K. G. Irvine, C. H. Carter, N. I. Kuznetsov, and E. V. Kalinina, “Electric breakdown in GaN P-n junctions,” Appl Phys. Lett., vol. 68, p. 229, 1996. [G] M. Ruff, H. Mitlehener, and R. Helbig, “SiC devices: physics and numerical simulation,” IEEE Trans. Electron Devices, vol. 41, pp. 1040–1054, June, 1994. [H] T. P. Chow and R. Tyagi, “Wide bandgap compound semiconductors for superior high-voltage unipolar power devices,” IEEE Trans. Electron Devices, vol. 41, pp. 1481–1483, Aug., 1994. [I] M. Bhatnagar and B. J. Baliga, “Comparison of 6H-SiC, 3C-SiC, and Si for power devices,” IEEE Trans. Electron Devices, vol. 40, pp. 645–655, Mar., 1993. between the old expressions and the newly derived ones are presented. The new expressions for specific on-resistance show that the use of wide bandgap semiconductor materials is much better than previously predicted. In addition, the maturity and expense of material processing will ultimately play a role in the optimal choice of the semiconductor system best suited for power semiconductor devices. It will be shown that the near-term optimal choice for uni-polar devices is clearly GaN. It will also be shown that the future optimal choice (qualified choice due to CTE mis-matches) for bipolar devices is C (diamond). II. OPTIMAL SEMICONDUCTOR MATERIALS In Table I various semiconductor material parameters are listed. The references for the material parameters in the table are denoted with capital letter superscripts and are listed separately in the reference section at the end of the paper.

From the table, it can be noted that the materials with a value of thermal conductivity close to or exceeding Si are GaN, GaP, SiC, and C (diamond). Of these four semiconductors, GaP has much lower carrier mobility values than Si. The best semiconductor material for the future is C (diamond). It has the largest thermal conductivity and bandgap of any of the materials from Table I. Diamond also has the largest electron mobility of any material from Table I with a bandgap larger than Si. However, there are two aspects of C (diamond) that make it less than ideal. First, the material and device fabrication technology is much less mature and developed than for SiC and GaN. Second, the CTE for C (diamond) is very low. Comparing the value of 0.8 ppm/K to typical package material CTE’s (listed in Table II and compared in Fig. 1) there is a clear thermo-mechanical mismatch, though this may be offset by the extremely high thermal conductivity and the future possibility of integrating diamond substrates directly with diamond-based semiconductor devices to eliminate the CTE mismatch. In fact,

HUDGINS et al.: WIDE BANDGAP SEMICONDUCTORS FOR POWER DEVICES

909

TABLE II PACKAGE MATERIAL PARAMETERS AT 300 K

Fig. 1. Coefficient of thermal expansion (CTE) for package substrates and some semiconductors (ppm/K). Note the close match between GaN and many package ceramics.

it could be reasoned that diamond devices on an insulating (intrinsic or polycrystalline) diamond substrate would cause no CTE mismatch and provide low thermal resistance and high electrical isolation. GaN and SiC are by comparison to C (diamond) very well suited to typical package materials, and in fact provide a better thermo-mechanical match than Si as shown in Fig. 1. The thermal conductivities of typical package materials and several semiconductors are graphed in Figs. 2 and 3, respectively. In addition, the material and device fabrication technology is much more advanced for SiC and GaN such that the near-term power device development will be directed into these material systems. Most discussions of packaging related to wide bandgap semiconductors have been with respect to operation at elevated temperatures. A good overview of general problems in using wide bandgap semiconductors and associated packaging at high temperatures is covered by [4]. There have been many reports of the thermal degradation of Schottky barriers and ohmic contacts in GaAs circuits and devices, and in AlGaN/GaN heterostructures, such as given by [5], [6]. Advanced electronic packaging materials and designs have been discussed thoroughly in [7]. Specific discussions of the ceramic substrate behavior, aging, and comparisons of use in module designs are described in [8]–[12]. SiC is beginning to be used in commercial power devices, particularly Schottky diodes, though the use of SiC in bipolar devices is questionable. However, GaN and related III–V devices are also being used in RF and electro-optic applications. Various poly-types of SiC shown below have electron mobilities

Fig. 2.

Thermal conductivity of common package materials (W/m 1 K).

Fig. 3. Thermal conductivity of semiconductors from Table I that are above or near the value of Si at 300 K (W/m 1 K). GaN* refers to GaN films grown over SiC substrates so that the thermal impedance of the device becomes closer to that of SiC.

similar, but generally less than in GaN. The hole mobilities are very similar in value to GaN, as well as the CTE values.

910

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 18, NO. 3, MAY 2003

Generally, maximizing the thermal conductivity of the semiconductor material and the associated package material is desirable. SiC has a larger thermal conductivity than GaN. However, it can be argued that this advantage is more than offset by the larger bandgap, and well known associated properties such as the theoretical maximum junction temperature of operation (intrinsic temperature) and the critical electric field, in GaN. It is therefore determined that GaN is at present, one of the best semiconductor materials for consideration in advanced power electronic devices based on a close CTE match to insulating ceramics, a reasonable value of electron mobility, and a thermal conductivity value not much different than in Si. III. CRITICAL ELECTRIC FIELD AND BANDGAP ENERGY The usual calculations of on-resistance based on the critical electric field of the material are shown to be inaccurate. Based on new calculations from additional data, a more precise on-resistance calculation shows even greater advantage to wider bandgap materials, thus favoring GaN over SiC further. A. Traditional Derivation The traditionally used empirical expression for ionization rates of holes and electrons is of the form: (1) where , , and are empirical fit parameters based on measured ionization rates of Ge, Si, GaAs, and GaP [13]. The funcis the position-dependent electric field associated with tion either abrupt junctions or linearly graded junctions. The expression in (1) then leads to a calculation of the critical electric field, , for breakdown in abrupt junctions. Using the expressions for abrupt junctions, a relationship between the critical field and bandgap can be created as given in [13], [14]: (2) where is the electron charge, is the permittivity of the semiis the impurity doping concentration conductor material, is the bandgap on the low-doped side of the junction, and energy. It should be noted that in Si power semiconductors the impurity concentration on the low-doped side of the junction is typically in the range of 10 –10 cm , resulting in a dependence of the critical field on the bandgap energy to the power . of 0.75 and relatively independent of the exact value of The expression in (2) has been used extensively in the literature [1]–[3], [14] to derive empirical dependencies between bandgap and critical field or breakdown voltage, and to fur, or specific on-resisther derive forms of on-resistance, , as figures of merit for power device performance tance, [1]–[3], [14]–[16]. B. New Derivation Using the values in Table I for critical field (in V/cm) and bandgap energy (in eV), a least squares fit to the data was per formed to derive a new relationship in the form of (3)

TABLE III PARAMETER VALUES

FOR RELATIONSHIP BETWEEN [REFERRED TO (3)]

E

AND

E

The form of (3) is proposed based upon the similarity to the form of (2). In (2), the critical field is empirically derived as being proportional to the bandgap energy raised to a power of 0.75 (given a fixed background doping concentration). A similar form is given by (3), but the leading coefficient and the exponent are derived based upon much more data than was used to derive (2). Equation (2) was derived using three indirect- and one direct-gap semiconductors combined. By comparison, the constants in (3) were derived using data for seven indirect- and six direct-gap semiconductors. The results of the derivation using the form of (3) are given below in Table III. Impurity doping dependencies were ignored due to the fact that power semiconductors typically utilize low doping concentrations to enhance voltage hold-off capabilities. It should be noted that the values in Table III indicate that there is great advantage to using direct-gap semiconductors to obtain large critical field values. As an optimal semiconductor for power electronics was determined to be GaN, from the previous discussions of thermo-mechanical properties above, the relationship between critical field and bandgap energy in direct gap materials is given below in (4). Comparison of (4) to (2) indicates a greater advantage to wide bandgap materials than was previously derived from the limited data set used by [13]. Using the numbers derived from Table III in (3) also allows estimation of unknown critical field values based solely on generally known bandgap values (4) For completeness, the relationship between critical field and bandgap energy for indirect gap semiconductors is given by (5) (5) Fig. 4(a) and (b) show a comparison of (2), (4), and (5) as well as the data from Table I, assuming a nominal base doping density of 10 cm . Equation (2) works well for Si and GaAs and somewhat less well for Ge. This is to be expected because (2) was derived by optimizing the relationship using only Si, Ge, GaAs, and GaP. Note that (2) greatly overestimates the critical field for small bandgap values and underestimates the critical field for wide bandgap materials as compared to the reported values. Use of (5) for indirect gap semiconductors provides a better fit to the experimental data reported in Table I than does (2). However, (5) does predict lower than reported values of critical electric field for 6H- and 4H-SiC. As the material quality of SiC improves, variations in reported values will reduce. It is expected that a future revision to (4) and (5) may

HUDGINS et al.: WIDE BANDGAP SEMICONDUCTORS FOR POWER DEVICES

911

terms of bandgap energy and concentration on the lightly doped side of the junction is given as (6) The corresponding expression based on the newly derived dependencies from (3) is given as (7) where and are the appropriate parameters from Table III, and is the relative permittivity of the material. Carrying through with the physical behavior associated with an abrupt junction , expresand using (3) and (7), a specific on-resistance, sion can be calculated as given in (8). The on-resistance value refers to the structure of a majority carrier device, specifically for electrons because of the larger value of mobility, and only accounting for the drift-region drop (e.g., ignoring other contributions to the device forward voltage drop during conduction) (a)

(8) Substituting specific parameter values of and , from Table III, . The reader is reinto (8) gives the final results, (9), for minded that the derived parameters are based on a power relation, (3), between the critical field and the energy bandgap using data from Table I. A least squares fit to the data is used to derive specific values of and Indirect Bandgap (9a) Direct Bandgap (9b) The new results in (9) are as compared to previous results given by [14]–[16]: (b) Fig. 4. (a) Comparison of (2) and (4) relating bandgap energy and critical electric field for direct gap semiconductors and associated reported data. (b) Comparison of (2) and (5) relating bandgap energy and critical electric field for indirect gap semiconductors and associated reported data.

be necessary as material quality in all of the semiconductors improves and values of the critical field stabilize for each material system. Note that (4) and (5) show the increase in critical field as the bandgap increases even when directly comparing directand indirect-gap semiconductors. An argument could be made that providing separate expressions is redundant, but based on a least-squares error fit to the data, the two expressions (4) and (5) present a better predictive tool than can be achieved using (2). More importantly, it should be stressed that (4) and (5) can be used primarily to make comparisons among material systems rather than making predictions about the absolute value of a particular critical field associated with a particular semiconductor. IV. SPECIFIC ON-RESISTANCE CALCULATION COMPARISON

AND

Using (1) and (2) as obtained from [13], and extended by [14], , for an abrupt junction in the voltage breakdown value,

(10) are compared in Fig. 5. Note that The equations for (10) overestimates the on-resistance for both SiC and GaN primarily because the basis for the derivation of (10) depends upon (2). Equation (2) was created based on data that was closely fitted to Si. The close overlay of curves in Fig. 5 for Si indicates the expected similar results. V. GaN AND RELATED MATERIALS, AND FABRICATED DEVICES Unique materials properties of GaN-based semiconductors have stimulated a great deal of interest in research and development in materials growth and opto-electronic, and electronic devices using this semiconductor system. The major advantages of nitride-based devices that make them extremely promising for high-power high-temperature applications are high electron mobility and saturation velocity, high sheet carrier concentration at heterojunction interfaces, high breakdown field, and low thermal impedance when grown over SiC substrates. The chemical inertness of nitrides is a key property to provide high reliability. The major problems with the nitride based devices are related to the material quality. Since there is no bulk GaN crystals avail-

912

Fig. 5.

R

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 18, NO. 3, MAY 2003

Comparison between (9a), (9b), and (10) for specific on-resistances, , as a function of breakdown voltage for Si, GaN, and SiC.

able to date, the epitaxial films are grown over sapphire or SiC substrates. In both cases due to significant lattice mismatching the films contain a high concentration of defects, which affects the breakdown voltage, the carrier mobility, device reliability and other important characteristics. Typical values for the room temperature electron mobility in the doped GaN films do not exceed 300 cm /V-s, though 900 cm /V-s has been reported (ref. Table I). Some Schottky rectifiers have been reported [17]. The breakthrough in the GaN based transistor development though, is associated with the demonstration of the high-density two-dimensional (2-D) electron gas at the AlGaN/GaN interface [18]. Due to extremely large bandgap offsets at the interface, the 2-D gas density in the AlGaN/GaN heterostructures can be as high as 2 10 cm , which is about 10 times higher compared to AlGaAs/GaAs structures. It was also shown that in the 2-D gas, the screening of the defect and impurity scattering allows for much higher mobility, up to 1500–2000 cm /V-s. Since the first demonstrations in 1991 and 1992, several groups had reported high power operation of AlGaN/GaN Heterostructure Field-Effect Transistors (HFET’s) at microwave frequencies [19]–[21], including a 100 W output power single chip amplifier developed by Cree, Inc. and 100 GHz cut-off frequency devices reported in [22]. The maximum current achievable in the nitride based , is 1–2 A/mm of the device periphery. HFET’s, – cm is the carrier sheet density Here – cm/s is at the AlGaN/GaN interface and the electron saturation velocity. The breakdown voltage of the HFET’s is limited either by the impact ionization in the channel or by the surface breakdown, depending on the device design and geometry. The device characteristics can further be improved by incorporating a thin, 100 Å, insulating layer under the gate. This insulating layer, which can be made using either SiO (MOSHFET) [23], or Si N (MISHFET) [24] films reduces the gate leakage current by 4–6 orders of magnitude and also allows for a large positive gate voltage swing. This almost doubles the

Fig. 6. (a) HFET device built using the GaN system. (b) MOSHFET device built using the GaN system with SiO under the gate.

Fig. 7. The breakdown voltage of MOSHFET as a function of gate-drain separation.

maximum current available from the device. Typical HFET and MOSHFET device structures are shown in the Fig. 6(a) and (b). The Fig. 7 shows the experimental dependence of the MOSHFET breakdown voltage on the gate-to-drain spacing. The measurements were done for the HFET’s with 1 m gate length fabricated over insulating SiC substrates. Two plots in Fig. 7 correspond to the HFET with an unprotected AlGaN

HUDGINS et al.: WIDE BANDGAP SEMICONDUCTORS FOR POWER DEVICES

Fig. 8. The I –V characteristics of the III–N MOSHFET based switch.

surface and to the device passivated with 0.3 m thick layer of silicon dioxide. The associated value of the electric field at breakdown in the MOSHFET is much lower than predicted by (3) and (4). The achieved breakdown, lower than the theoretical maximum value of critical field, is currently limited because of the lack of material uniformity and the large number of crystal lattice defects. It is expected that the actual breakdown will approach the theoretical limit as device processing and material growth technology mature. Fig. 8 shows the – characteristics of the AlGaN/GaN MOSHFET’s for high-power switching applications [25]. In the figure, the device current density was normalized to the total device area, which includes the 5 m source-drain spacing and 50 m wide source and drain ohmic contacts. As seen, the combination of high current density and large breakdown voltage make MOSHFET’s promising candidates for high power switching applications. The on-state specific resistance is seen to be 75 m -mm (7.5 10 -cm ) from Fig. 8. In comparison, the predicted theoretical specific on-resistance for a 500 V breakdown device -cm as shown in Fig. 5. (9b) built in GaN is around 2 10 The previously used relationship for specific on-resistance (10), also predicts a lower than realized value. These results indicate that the material quality problems associated with GaN and AlGaN heterostructures has considerable room for improvement left before truly high quality power devices can be realized. However, these improvements are progressing at a rapid pace and much improved device performance can be expected in the next few years. The MOSHFET is a normally-on device and thus requires active gating to keep it off. This can be a limitation in some power electronic circuits, however, many gate-drivers for Si IGBT’s, MOSFET’s, and thyristors often use a small reverse bias applied to the gate to help with noise immunity and blocking capability. It is thus noted that an active gate signal during the off-state of the MOSHFET is not a particular handicap. Generally, a vertical structure is preferred in power devices to limit surface effects on the bulk device performance and to maximize current density capability. Improvements

913

in material quality for GaN/AlGaN and SiC systems will greatly enhance performance of planar structures. Ultimately, vertical versions of devices such as MOSHFET’s will improve performance further. A suitable package design for operating devices designed using GaN and related materials, such as the MOSHFET, would include pressure contacts to eliminate all wirebond connections and solder layers. A substrate material of interest is AlSiC because of the close CTE match to GaN and the fact that many GaN devices are grown on SiC substrates. The use of AlSiC/SiC in direct contact could provide a good starting platform for device growth. Another ceramic of interest is using polycrystalline AlN with the GaN/AlGaN device grown directly on top. This would provide a very good CTE match (between molar fractions of AlGaN in the electronic device and AlN in the insulating substrate) between materials. The envisioned package would be very desirable if operating conditions required high junction temperatures. VI. CONCLUSION It was argued that GaN, SiC, and C (diamond) are the best semiconductor material systems in which to create future power electronic devices. Very few semiconductors have an electron to hole mobility ratio below three and also have an electron mobility above 1000 cm /V s. Diamond is one of the few that meets these conditions. The near unity mobility ratio of diamond makes it ideal for bipolar device designs, particularly in operating environments of elevated temperatures. The close match and high values of carrier mobilities, as well as the large bandgap and high thermal conductivity, make diamond the ideal future material for electronic devices of all power levels and types. However, in light of the maturity of the fabrication technology, large bandgap (larger than SiC) and thermo-mechanical material properties, GaN appears to be the next best choice, overall, for the upcoming decade of development. Only slightly less desirable is the use of SiC for power electronic devices. The material quality and impurity doping issues associated with SiC make it presently unusable for high powered bipolar devices. However, because 4H-SiC has a bandgap close to GaN, the material system that can address the quality issues first and best will likely win out. It was shown from the specific on-resistance comparison that theoretically GaN should be a better choice for optimizing electrical behavior, though it and SiC are much better than using Si. It has been shown from data reported over the past decade that the traditionally used relationship between energy bandgap and critical electric field is incorrect to use with semiconductor materials for which it was not optimized. Two new relationships, one for direct-gap and another for indirect-gap materials have been derived using reported data for 13 different semiconductors. A least-squares fit to the data was performed and coefficients derived for an exponential relation. The associated relationships for specific on-resistance were re-calculated in light of the new bandgap-critical field equations. A comparison of the was given and showed that wide revised equations for bandgap materials are potentially better for high voltage devices than previously thought.

914

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 18, NO. 3, MAY 2003

Finally, a discussion of some practical material concerns for GaN and related semiconductor systems was provided. It was shown that reduced carrier mobility could be overcome by 2-D electron gas formation using heterojunction structures. Specific examples for a MOS device was given (MOSHFET) that incorporated layers of AlGaN, GaN, and AlN on both SiC and sapphire substrates.

REFERENCES [1] T. P. Chow and R. Tyagi, “Wide bandgap compound semiconductors for superior high-voltage unipolar power devices,” IEEE Trans. Electron Devices, vol. 41, pp. 1481–1483, Aug. 1994. [2] M. Bhatnagar and B. J. Baliga, “Comparison of 6H-SiC, 3C-SiC, and Si for power devices,” IEEE Trans. Electron Devices, vol. 40, pp. 645–655, Mar. 1993. [3] B. J. Baliga, “Power semiconductor device figure of merit for high-frequency applications,” IEEE Electron Device Lett., vol. 10, pp. 455–457, Oct. 1989. [4] F. P. McClusky, R. Grzybowski, and T. Podlesak, Eds., High Temperature Electronics. Orlando, FL: CRC Press, 1997. [5] E. K. Braun, K. V. Shenoy, C. G. Fonstad, and J. M. Mikkelson, “Elevated temperature stability of GaAs digital integrated circuits,” IEEE Electron Device Lett., vol. 17, pp. 37–39, Feb. 1996. [6] I. Daumiller, C. Kirchner, M. Kamp, K. J. Ebeling, and E. Kohn, “Evaluation of the temperature stability of AlGaN/GaN heterostructure FETs,” IEEE Electron Device Lett., vol. 20, pp. 448–450, Sept. 1999. [7] J. E. Morris, Ed., Electronics Packaging Forum. New York: Van Nostrand Reinhold, 1991, vol. 2. [8] S. Catellani, J. C. Crebier, C. Schaeffer, and T. Marsala, “Thermal and electrical aging of DBC substrates,” in Proc. IEEE PESC’01, Vancouver, BC, Canada, June 2001. [9] E. S. Dettmer, B. M. Romenesko, H. K. Charles, B. G. Carkhuff, and D. J. Merrill, “Steady-state thermal conductivity measurements of AlN and SiC substrate materials,” IEEE Trans. Comp., Hybrids, Manufact. Technol., vol. 12, pp. 543–547, Dec. 1989. [10] G. Mitic, H. P. Degischer, G. Lefranc, and T. Licht, “AlSiC composite materials in IGBT power modules,” in Proc. PCIM’00 Conf., 2000. [11] C. V. Godbold, V. A. Sankaran, and J. L. Hudgins, “Novel designs in power modules,” in Proc. IEEE IAS Annu. Mtg., Orlando, FL, Oct. 1995, pp. 911–915. [12] C. V. Godbold, V. A. Sankaran, and J. L. Hudgins, “Thermal analysis of high-power modules,” IEEE Trans. Power Electronics, vol. 12, pp. 3–11, Jan. 1997. [13] S. M. Sze and G. Gibbons, “Avalanche breakdown voltages of abrupt and linearly graded junctions in Ge, Si, GaAs, and GaP,” Appl. Phys. Lett., vol. 8, no. 5, pp. 111–113, Mar. 1966. [14] B. J. Baliga, “Semiconductors for high-voltage, vertical channel field effect transistors,” J. Appl. Phys., vol. 53, no. 3, pp. 1759–1764, Mar. 1982. [15] B. J. Baliga, Power Semiconductor Devices. New York: PWS Publishing Co., 1996. [16] A. Q. Huang and B. Zhang, “The future of bipolar power transistors,” IEEE Trans. Electron Devices, vol. 48, pp. 2535–2543, Nov. 2001. [17] G. T. Dang, A. P. Zhang, F. Ren, X. A. Cao, S. J. Pearton, H. Cho, J. Han, J. Chyi, C. M. Lee, C. C. Chuo, S. N. Chu, and R. G. Wilson, “High voltage GaN Schottky rectifiers,” IEEE Trans. Electron Devices, vol. 47, pp. 692–695, Apr. 2000. [18] M. A. Khan, J. N. Kuznia, J. M. Van Hove, N. Pan, and N. Carter, “Observation of a two-dimensional electron gas in low pressure MOCVD deposited GaN/AlGaN heterojunctions,” Appl. Phys. Lett., vol. 60, pp. 3027–, 1992. [19] T. Sheppard, W. L. Pribble, D. T. Emerson, Z. Ring, R. P. Smith, S. T. Allen, and J. W. Palmour, Proc. Device Res. Conf. Dig., 2000, p. 37. [20] N. X. Nguyen, M. Micovic, W.-S. Wong, P. Hashimoto, L.-M. McCray, P. Janke, and C. Nguyen, Electron. Lett., vol. 36, pp. 468–, 2000. [21] Y.-F. Wu, D. Kapolnek, J. P. Ibbetson, P. Parikh, B. P. Keller, and U. K. Mishra, “Very-high power density AlGaN/GaN HEMTs,” IEEE Trans. Electron Devices, vol. 48, pp. 586–590, 2001. [22] W. Lu, J. Yang, M. A. Khan, and I. Adesida, “AlGaN/GaN HEMT’s on SiC with over 100 GHz f and low microwave noise,” IEEE Trans. Electron Devices, vol. 48, pp. 581–585, 2001.

[23] M. Asif Khan, X. Hu, G. Simin, A. Lunev, J. Yang, R. Gaska, and M. S. Shur, “AlGaN/GaN metal–oxide-semiconductor heterostructure field effect transistor,” IEEE Electron Device Lett., vol. 21, pp. 63–65, Feb. 2000. [24] X. Hu, A. Koudymov, G. Simin, J. Yang, M. Asif Khan, A. Tarakji, M. S. Shur, and R. Gaska, “Si3N4/AlGaN/GaN-metal–insulator-semiconductor heterostructure field effect transistors,” Appl. Phys. Lett., vol. 79, pp. 2832–2834, 2001. [25] G. Simin, X. Hu, N. Ilinskaya, A. Kumar, A. Koudymov, J. Zhang, M. Asif Khan, R. Gaska, and M. S. Shur, “A 7.5 kW/mm2 current switch using AlGaN/GaN metal–oxide-semiconductor heterostructure field effect transistors on SiC substrates,” Electron. Lett., vol. 36, pp. 2043–2044, 2000. Jerry L. Hudgins (SM’91) received the Ph.D. degree in electrical engineering from Texas Tech University, Lubbock, TX, in 1985. He is presently a Litman Distinguished Professor of Engineering in the Electrical Engineering Department, University of South Carolina, Columbia, where he served as Interim Department Chair, from 1998 to 2000. He has published over 50 technical papers and book chapters concerning power semiconductors and engineering education, and has worked with numerous industries. Dr. Hudgins served as the President of the IEEE Power Electronics Society (PELS), in 1997 and 1998. He is President of the IEEE Industry Applications Society (IAS) for 2003. Grigory Simin (M’01) received the M.S.S.E degree from Leningrad Electrotechnical Institute, Leningrad, Russia, in 1971, the Ph.D. degree in the physics of semiconductors and dielectrics from GIRICOND Science and Research Institute, Leningrad, Russia, and the Senior Research Scientist Diploma from the Supreme State Committee on Academic Degrees, Russia, in 1985. During this period, his research has included Gunn effect devices, GaAs MESFET’s, microwave and optical integrated circuits, semiconductor lasers. Since 1998, he has been s a Research Professor and Associate Professor (since 2001) with the Department of Electrical Engineering, University of South Carolina, Columbia. His research focus is GaN based electronic and optoelectronic devices. He published four books, several book chapters on the physics of semiconductor devices, and more than 100 articles in referred journals. Enrico Santi (SM’01) received the Dr.Ing. degree in electrical engineering from the University of Padua, Italy, in 1988 , and the M.S. and Ph.D. degrees from the California Institute of Technology, Pasadena, in 1989 and 1994, respectively. He was a Senior Design Engineer at TESLAco, from 1993 to 1998, where he was responsible for the development of various switching power supplies for commercial applications. Since 1998, he has been an Assistant Professor in the Electrical Engineering Eepartment, University of South Carolina, Columbia. He has published several papers in power electronics and holds two patents. His research interests include switching-mode power converters, modeling of power semiconductor devices, and simulation of advanced power systems. M. Asif Khan (M’81) received the Ph.D. degree from the Massachusetts Institute of Technology, Cambridge, in 1979. He joined the University of South Carolina to head up the new Microelectronics Laboratory. The initial focus of the program is on the fabrication of AlGaN and SiC based high power microwave transistors for high temperature operations. He was previously Vice President, Optoelectronics, APA Optics, Inc., Blaine, MN, where his research group did the pioneering work in the development of GaN-AlGaN materials and devices. Prior to that, he worked as the principal scientist in research and development groups at 3M and Honeywell, Inc. He has authored over 150 refereed papers, several book chapters, and over 25 invited papers.