Wide gap semiconductor microwave devices

IOP PUBLISHING JOURNAL OF PHYSICS D: APPLIED PHYSICS J. Phys. D: Appl. Phys. 40 (2007) 6355–6385 doi:10.1088/0022-3727/40/20/S18 Wide gap semiconductor microwave devices V V Buniatyan1 and V M Aroutiounian2 1 Armenian State Engineering University, 105 Teryan Street, 375009, Yerevan, Armenia Department of Physics of Semiconductors and Microelectronics, Yerevan State University, 1 Manoukian Street, Yerevan 375025, Armenia 2 Received 29 May 2007 Published 5 October 2007 Online at stacks.iop.org/JPhysD/40/6355 Abstract A review of properties of wide gap semiconductor materials such as diamond, diamond-like carbon films, SiC, GaP, GaN and AlGaN/GaN that are relevant to electronic, optoelectronic and microwave applications is presented. We discuss the latest situation and perspectives based on experimental and theoretical results obtained for wide gap semiconductor devices. Parameters are taken from the literature and from some of our theoretical works. The correspondence between theoretical results and parameters of devices is critically analysed. (Some figures in this article are in colour only in the electronic version) Introduction 1. Material technology and properties The recognition of the potential usefulness of wide bandgap (WBG) semiconductor materials for high-frequency, highpower, high-temperature and harsh environment applications started over 20 years ago [1–10]. However, for various reasons related to both device electronics and materials technology, intensive development efforts have been made in this direction only over the last decade. Therefore the WBG semiconductor technology is today less developed than the mature siliconbased semiconductor technology. WBG materials and devices are less reproducible and more expensive compared with silicon ones. Anyway, the main driving force for this resurgent interest is the following: (1) devices and systems made of WBG materials will become widely available on a large variety of automotive, aerospace, deep-well drilling, shipboard and other industrial systems; (2) these materials are promising candidates for blue and ultraviolet light emitting diodes (LED) and lasers; (3) The Keyes and Johnson figures of merit are considerably higher for these materials as compared with those of Si and GaAs; (4) WBG devices used in low-power transistors at ambient temperatures exceeding 300 ◦ C are commercially available. Silicon and silicon-on-insulator devices already satisfy the requirements for very large-scale digital and analog integrated circuits operating in lower temperature range. The possibilities of using different semiconductors for harsh environment applications are shown in table 1 [7]. A number of review papers [1,2,4,7] recently appeared covering different aspects of WBG materials and their other applications. The expected excellent performances of WBG devices are often expressed by figures of merit. Keyes’ figure of merit (FOM) takes into account the switching speed of transistors and its thermal limitation due to generated heat that must be removed. Johnson’s figure of merit considers the high-frequency and high-power capability of devices. Baliga’s figure of merit concerns the on-resistance of power MOSFETs and is used to evaluate quantitatively the superiority of a device. The Keys, Johnson and Baliga figures of merit for some WBGbased devices are shown in table 2 [8, 11]. Following the Baliga FOM for high-frequency applications, only diamond is still superior to SiC. When ambient temperature increases, semiconductor device characteristics degrade until they no longer provide sufficient functionality for desired circuit applications. The fundamental physical limitations of silicon operation are perhaps the strongest motivations for the development of hightemperature WBG semiconductors. A number of factors are proposed that describe limitations on the high-temperature operation of semiconductor electronic devices and circuits. 0022-3727/07/206355+31$30.00 © 2007 IOP Publishing Ltd Printed in the UK 6355 V V Buniatyan and V M Aroutiounian Table 1. Possible semiconductor substrates for harsh environment applications [7]. Substrate Bandgap (eV) Electronic maximum operating temperature (◦ C) Processing maturity Key technical issues and limitations Si 1.1 150 350 Si: very high SOI: medium High 2.39 600 Low 6H–SiC 3.02 700 Low 4H–SiC 3.26 750 Low Nitrides 3.45–6.20 >750 Very low Diamond 5.48 1100 Very low Not suitable for harsh environments Contact stability at high temperature not suitable for aggressive environments Not available as bulk material Bulk material quality ohmic contacts to p-type material Bulk material quality ohmic contacts to p-type material Material quality ohmic contacts n-Type doping only polycrystalline material available GaAs 1.43 3C–SiC Table 2. Figures of merits for various semiconductors [11]. a Material JM (Ec vsat /π )2 KMb λ(vsat /εr )1/2 QF1c g λσA QF2d λσA EC Si GaAs 4H–SiC GaN GaP Diamond AlN 1 7.1 278 756 37 2540 5120 1 0.45 5.1 1.6 0.7 32.1 21 1 5.2 594 560 10 54.860 52.890 1 6.9 4357 6220 40 1.024 000 2.059 000 BMe σA = εr µEc3 1 16 178 650 16 4110 31 700 BHMf µEc2 1 11 29 77.8 5 470 1100 Note: All values are normalized with respect to Si. Johnson’s FOM for basic limit on the device performance (high power and frequency). Keyes’s FOM based on the switching speed of transistor. c Quality factor 1 (thermal FOM) for heat sink material and the active device area in power device. d Quality factor 2 is based on the perfect heat sink. e Quality factor 3 is based on any assumptions about materials or geometry. f Baliga FOM for evaluation of high frequency application. a b One of the temperature degradation mechanisms relates to the concentration of free carriers that governs device operation. As is well known, the concentration of intrinsic carriers (ni cm−3 ) is exponentially dependent upon the temperature of the semiconductor [12]. Calculated temperature dependence of intrinsic carrier concentration ni (T ) is shown in figure 1 for silicon, 6H–SIC and 2H–GaN. Some of the known high-temperature electronics application areas of WBGs are illustrated in table 3 [4]. At room temperature, intrinsic concentration of electrons ni for silicon (bandgap 1.1 eV) is around 1010 cm−3 , which is negligible compared with the 1014 –1018 typical for usual device doping levels. At higher ambient temperatures, the concentration of intrinsic charge carriers can be, by order of magnitude, the same as or more than the concentration of impurity carriers and govern the properties of a device instead of designed doping. This is easy to see in figure 1(b). Therefore, the situation mentioned with intrinsic and impurity concentrations is not realized in WBG semiconductors (for example SiC, GaN) up to much higher temperatures (600 ◦ C). Perhaps, the difference in intrinsic carrier concentration between silicon and WBG semiconductors becomes even more important for the current– voltage (I –V ) curve for rectifying junctions in semiconductor devices. As WBGs have much lower intrinsic carrier 6356 concentrations than silicon, the leakage current is orders of magnitude lower than in the case of silicon. Therefore, WBG devices are capable of operating at much higher temperatures than the fundamental limitation of 600 ◦ C. Another fundamental leakage mechanism in devices is the transport (emission) of carriers having sufficient energy to go over or tunnel through an energy barrier in the device structure. Moreover, for the field emission model, based on the Padovani and Stratton expression [12], for a low temperature or for a high applied field, a simple expression for field emission current density IFE as a function of the electric field E is derived [13] under the condition that the tunnelling current from the semiconductor can be disregarded. The carrier emission leakage current for the thermal ionic (thermionic) field emission model is exponentially reduced when the effective potential height φB is increased. In the intermediate electric field region, thermionic field emission current is dominant (figure 2). The crossover point of the thermionic field emission and the tunnelling field emission goes up as the Schottky barrier height (SBH) increases. If the temperature increases, the boundary of the field emission region and the thermionic field emission region shifts towards higher electric fields because the temperature dependence of the field emission current is weaker than that of the thermionic field emission current. Figures 3 and 4 show calculated Wide gap semiconductor microwave devices Table 3. Semiconductor technologies for some selected high-temperature electronics [4]. High-temperature electronics application (a) Automotive Engine control electronics On-cylinder and exhaust pipe Electric suspension and brakes Electric/hybrid vehicle PMAD Turbine engine Sensors, telemetry, control reverse characteristics of a 4H–SiC and Si Schottky barrier diode (SBD). In the reverse characteristics of a Si SBD, the increase in the leakage current induced by barrier lowering is also important, and the effect of the tunnelling process on the leakage current is relatively small in comparison with that for a SiC SBD, because the electric field applied to the metal–semiconductor interface is much lower than that in SiC SBD. An important technological problem presently limiting the performance of WBG devices is in some instances the fabrication of high-temperature stable low-contact resistance ohmic contacts. It would also be desirable to be able to fabricate highly perfect Schottky barrier metal contacts on WBG semiconductors. Defect free, sharp epitaxial interfaces are desirable for reducing the resistance of ohmic contacts as well as increasing the perfectness factor of Schottky barriers. Realistic Schottky barrier (SB) contacts for most semiconductors have effective barrier heights usually less than three quarters of the bandgap energy. Thus, while silicon SBHs (with 1.1 eV bandgap) are practically confined to less than 0.9 V, twice or higher BHs can be realized in WBG semiconductors (∼3 eV), which reduces the leakage current at any given temperature by at least several orders of magnitude. This enables the realization of high-voltage and high-temperature WBG Schottky power rectifiers, which may not be realized using conventional semiconductors. Above Future technology 150 ◦ C <1 kW 600 ◦ C <1 kW BS and SOI NA BS and SOI WBG 250 ◦ C >10 kW BS WBG 150 ◦ C >10 kW BS WBG 300 ◦ C <1 kW SOI and WBG WBG WBG 600 ◦ C 150 ◦ C 300 ◦ C 550 ◦ C >1 kW BS and SOI WBG >10 kW NA WBG ∼1 kW NA WBG 300 ◦ C <1 kW SOI SOI 600 ◦ C <1 kW NA WBG Deep-well drilling telemetry Oil and gas 300 ◦ C <1 kW SOI Geothermal 600 ◦ C <1 kW NA SD1 and WBQ WBG Spacecraft Power management Venus and Mercury exploration Industrial High-temperature processing Figure 1. (a) Semiconductor intrinsic carrier concentration (ni ) versus temperature for silicon, 6H–SiC, and 2H–GaN. (b) Intrinsic carrier concentration for various semiconductors. Current technology BS and SOI <1 kW NA >10 kW BS and SOI >10 kW NA Electric actuation (b) Peak Chip ambient power 600 ◦ C 150 ◦ C WBG Note: BS—bulk silicon, SOI—SOI, NA—not presently available, WBG—wide bandgap. room temperature the ability of carriers to move through a semiconductor crystal decreases. This decrease in carrier mobility with the temperature, due to lattice scattering, reduces the value of current when a diode or transistor is manufactured. In general, the increase in the semiconductor device resistance with the temperature follows a T X power law, where x is usually between 1.5 and 2.5 above room temperature in most of the semiconductors mentioned, where lattice scattering is dominant (including silicon and WBG semiconductors [12]). Like Si, the lattice scattering (acoustic phonons) and ionized impurity scattering, together with (anisotropic) piezoelectric stuttering [14], seem to be the most relevant mechanisms to limit the mean free path of charge carriers at low electric fields in WBG semiconductors. For example, for SiC and GaN, the dependence of charge carrier mobility on the temperature and the dopant concentration may be expressed as [15, 16] Bi (N )(T /T0 )βi , 1 + Bi (N )(T /T0 )αi +βi (1)  µmin,i + µmin,i (Ngi /N )γi  .  µmax,i − µmin,i T =T0 =300 K (2) µi (N, T ) = µmax i (T0 ) where Bi (N ) =
6357 V V Buniatyan and V M Aroutiounian Table 4. The values of the constants µmax i , µmin i , i, Ngi and γi providing the best fit to the experimental data [15]. Type of carrier µmax i µmin i Ngi Polytypes (cm2 V−1 s−1 ) (cm2 V−1 s−1 ) (cm−3 ) Electrons 4H–SiC 6H–SiC 3C–SiC Holes Figure 2. The dominant mechanism of the reverse leak current of the 4H–SiC SBD. 4H–SiC 6H–SiC 30 40 20 25 30 50 — 33 5 25 2 × 10 2 × 1017 4.5 × 1017 1018 6 × 1017 2 × 1017 3 × 1016 1019 1019 5 × 1018 0.67 0.76 0.45 0.8 0.8 0.8 0.8 0.5 0.5 0.4 Table 5. The values of the constants µmax i , µmin i , i, Ngi and γi providing the best fit to the experimental data [16]. Type of carrier Tunnelling Model 880 950 400 400 420 750 650 117 75 95 γi 17 Electrons Holes µmax i (cm2 V−1 s−1 ) 1000 170 µmin i (cm2 V−1 s−1 ) Ngi (cm−3 ) γi 55 3 2 × 10 3 × 1017 1.0 2.0 17 Figure 3. Calculated reverse characteristics of 4H–SiC SBD. Tunnelling Model Figure 5. Temperature dependences of the low-field electron mobility in 4H–SiC (for F ⊥ c-axis, T = 300 K): curve 1—N = 5 × 1014 cm−3 ; curve 2—N = 4 × 1016 cm−3 . Figure 4. Calculated reverse characteristics of Si SBD. Here I is equal to 1 or 2 for electrons or holes, respectively. The values of the constants µmax i , µmin i , i, Ngi and γi are given in table 4 for SiC and in table 5 for GaN [15, 16]. Figures 5–7 present the temperature dependences of the mobility in a wide temperature range for 4N–SiC, 6H–SiC, 3C-SiC and in figures 8–10 for GaN. The values αi and βi providing the best fit to the experimental data for SiC and GaN are given in tables 6 and 7, respectively [15, 16]. 2. SiC-based microwave devices High breakdown-electric field strength, rather high electron mobility, wide bandgap, high thermal conductivity, highelectron saturation velocity and high-radiation stability make silicon carbide (SiC) an attractive candidate for fabrication of high-temperature, high-frequency and high-power electronic 6358 Figure 6. Temperature dependences of the low-field electron mobility in 6H–SiC (for F ⊥ c-axis, T = 300 K) for various values of N (cm−3 ): curve 1—5 × 1015 cm−3 ; curve 2—2 × 1016 cm−3 ; curve 3—8 × 1016 cm−3 ; curve 4—2 × 1017 cm−3 ; curve 5—2 × 1018 cm−3 . Wide gap semiconductor microwave devices Figure 7. Temperature dependences of the low-field electron mobility in 3C–SiC, at T = 300 K, for various values of N (cm−3 ): curve 1—3 × 1016 cm−3 ; curve 2—4 × 1017 cm−3 ; curve 3—1 × 1018 cm−3 ; curve 4—2 × 1019 cm−3 . Figure 10. Temperature dependence of the low-field electron mobility in wurtzite GaN at different values of dopant concentration. Curves represent the best fit according to equations (1) and (2) and parameter values listed in table 5: 1—N = 3 × 1016 cm−3 , 2—N = 1017 cm−3 , 3—N = 1.5 × 1017 cm−3 , 4—N = 2 × 1017 cm−3 , 5—N = 3.5 × 1017 cm−3 , 6—N = 1018 cm−3 , 7—N = 3 × 1018 cm−3 . Experimental data are taken from [5]. Table 6. The values of αi and βi providing the best fit to the experimental data for 4H–, 6H– and 3C–SiC [15]. Figure 8. The low-field hole mobility as a function of concentration of impurities in wurtzite GaN at room temperature. The curve represents the best fit according to equation (5) and data presented in table 5. Type of carriers Polytype αi βi Electrons 4H–SiC 6H–SiC 3C–SiC 4H–SiC 6H–SiC 3C–SiC 2.6 2.1 2.5 — 2.0 2.2 0.5 0.7 0.3 — — — Holes Table 7. The values of the parameters αi and βi providing the best fit to the experimental data for GaN [16]. Figure 9. The low-field electron mobility as a function of dopant concentration in GaN at room temperature. The curve represents the best fit according to equation (1) and data presented in table 5. devices [1–6]. The growth and processing techniques for SiC have been improved tremendously over the past years. As material quality and device processing in SiC continue to mature rapidly, it is now possible to fabricate various Type of carriers αi βi Electrons Holes 2.0 5.0 0.7 — new semiconductor devices made of SiC single crystals and epitaxial layers. On the other hand, during different technological processes, deep levels and trap centres with different concentrations and energy distributions can be formed and open the possibility of varying the characteristics of SiC devices. Recent advances in crystal growth and thin film epitaxy allow the development of high-quality structures made of different SiC polytypes [1, 3, 7, 17–26] (see table 8 [5]). Note also that unlike other WBGS, high quality, virtually defect-free large wafer is now available for silicon carbide [139]. Rather small lattice SiC mismatches with AlN, GaN, Si, as well as the abundance of polytypes in SiC, make SiC a material with immense potential for the manufacture of different heterojunction electronic devices using differing bandgap, charge carrier mobility, etc. SiC p-i-n rectifiers and other power electronic devices with good dc performance have been commercially available for many years (for a summary 6359 V V Buniatyan and V M Aroutiounian Table 8. Silicon carbide material properties [5]. Quantity 3C–SiC 4H–SiC 6H–SiC Si Eg (eV) at T < 5 K Ecrit (MV cm−1 ) Ecrit (ND ) (kV cm−1 ) K(W cm−1 K−1 ) at 300 Ka ni (cm−3 ) at 300 Kb vsat (cm s−1 ) (parallel to c-axis) µe (cm2 V−1 s−1 ) m⊥ /m|| at 300 K µh (cm2 V−1 s−1 ) χ (eV) at 300 K Refractive index (n) εS 2.40 3.26 3.02 1.12 2.12 2.2 2.5 0.25 3.7 10.64 · ND 0.142 4.9 1.5 a b 3.2 1.5 × 10−1 5 × 10−9 1.6 × 10−6 1.0 × 1010 800 2.0 × 107 2.0 × 107 1.0 × 107 1000 1400 400 0.7–0.83 6 40 2.7 9.72 115 101 471 2.70 2.95 3.15 2.712 2.7 3.5 9.66 11.7 −3 Doped at ∼10 cm . Nc , Nv ∼ 1019 cm−3 . 17 of high-voltage SiC devices, see in [4]). Additionally, devices such as heterojunction field effect transistor (HFET) and heterojunction bipolar transistor (HBT) have been proposed and fabricated [3, 4, 17–29]. WBG semiconductors SiC and GaN have been viewed as highly promising for microwave power generators, amplifiers and other applications. The advantages of WBG materials over conventional Si and GaAs include high breakdown field, high saturation electron velocity and high thermal conductivity. The output power of microwave power devices made of WBG semiconductor materials is higher than that of conventional microwave power devices made of Si or GaAs. Although SiC substrates offer a thermal conductivity of 4 W cm−1 K−1 compared with 1.5 W cm−1 K−1 of Si and 0.5 W cm−1 K−1 of GaAs, the potentials of WBG device performance will be significantly compromised if heat dissipation is not properly managed. For both SiC MESFETs and GaN/AlGaN HEMTs, the full channel current at 300 ◦ C is only about 55% of that at room temperature mainly due to the electron mobility reduction at higher temperatures. So, a careful simulation of WBG devices is necessary [27]. The FOM of unipolar devices made of 4H–SiC can be, theoretically, three orders of magnitude higher than that for their silicon counterparts. FOM as high as 485 MW cm−2 , 827 MW cm−2 and 1727 MW cm−2 have been reported for 700 V 4H–SiC JFET, 1.7 kV 4H–SiC JFET and 9.2 kV 4H–SiC BJT [29]. These FOMs are, respectively, 97, 165 and 345 times higher than the theoretical unipolar limit for Si. Many RF and high-frequency devices made of SiC are also commercially available or rapidly approaching the commercialization state [1, 3, 17]. In particular, metal semiconductor field effect transistors (MESFETs) and static injection transistors (SIT) [21–23] with excellent dc and RF 6360 performance have been demonstrated and these devices are being developed for microwave power amplifier and oscillator applications. For SiC MESFETs, 250 µm periphery devices have demonstrated a record power density of 5.6 W mm−1 at 3 GHz [11, 30] and 5.2 W mm−1 at 9.5 GHz [21]. The largest SiC MESFETs had 48 mm gate periphery with 80 W CW power (1.67 W mm−1 ) at 3 GHz [30]. For SiC MESFETs operating at lower frequencies, small gate periphery devices were reported [31, 32], but no large gate periphery devices have been reported for UHF band applications [32]. SiC MESFETs operating at UHF band (450 MHz) with 62 W output power have been realized in [32] from single 21.6 mm gate periphery devices, which is the highest power from a single device at UHF band. SiC MESFETs operating at 3 GHz have delivered 27 W output power from a single 14.4 mm gate periphery device. Surface passivation is needed for SiC MESFETs to eliminate the surface trapping effects. A 4H–SiC MESFET structure had output power of 6.2 W mm−1 at 1 GHz [32]. Cree Inc. is marketing MESFETs with a best performance of 5.2 W mm−1 at 3.5 GHz and 63% power added efficiency (PAE) for a 0.7 µm × 48 mm device. Monolithic microwave IC (MMIC) amplifiers based on these devices are also demonstrated with a power of 36.3 W, associated gain of 0.5 dB and PAE of 20.6% [11]. The noise characteristics of SiC MESFETs are calculated in [33] after detailed analysis of physical processes taking place in the small-signal regime of operation. Power MOSFETs are useful for high-frequency applications due to their high inherent switching speed. Application of Si power MOSFETs has been limited to circuits with voltages of up to 1000 V at current levels of only 1 A. This is mainly because of high specific on-resistance associated with Si MOSFETs. For applications at above 200 V, IGBTs with high-end current ratings of 1500 V/300 A are preferred today. As the on-resistance of SiC MOSFET is at least two orders of magnitude lower than that for Si MOSFET, it makes SiC MOSFET an attractive alternative device for existing Si power devices (including GTOs and thyristors) [5, 2]. The analysis suggests that SiC power rectifiers and MOSFETs could be a superior alternative to all Si power devices with breakdown voltage of as high as 5000 V. However, this would require considerable improvements in the SiC device processing technology. The role of different type transit-time diodes in microwave applications is well known. In general, the use of impact-avalanche-transit-time (IMPATT) diodes [34, 35] in the millimeter and sub-millimeter ranges of wavelengths is associated with serious difficulties. The avalanche-induced dispersion effect, time delay and carrier diffusion effects in the IMPATTs limit their high-frequency operation and increase the noise level. Another well-known transit-time device is the barrier injection transit-time (BARITT) diode [36–39], where charge carriers are injected in transit-time space thermally over the potential barrier. Low drift velocity and high diffusivity around the injection point limit high-frequency performance of BARITTs. As a rule, because of the unfavourable injection phase delay, the BARITT diode operates at lower power and efficiency in comparison with the IMPATT diode, but BARITTs have remarkably low noise. When the thickness of the IMPATT diode becomes narrower (in order to increase the oscillation frequency), avalanche mechanisms weaken and Wide gap semiconductor microwave devices the tunnel injection becomes dominant in reverse biased p–n junctions. This, in turn, causes corresponding changes in microwave characteristics of diodes [40–44]. For example, if the characteristic frequency of the tunnelling ωt is defined as  2q 2 ωt = · Et
ω; (3) m∗ Eg the tunnelling can be assumed to be non-inertial. Here Et is the average electric field, which is determined by the ‘height’ of the tunnelling barrier and effective mass of charge carriers m∗ . Eg is the width of the bandgap. To reduce the noise and decrease the intrinsic response time in the injection region, the tunnelling is used as a desirable injection mechanism. The non-inertial tunnelling phenomenon is less preferable for the microwave generation of power than impact ionization as the phase delay in tunnelling structures can be realized at the sacrifice of transit-time effects. This leads to lower efficiency and output power in tunnel transit-time diodes (TUNNETTs). But higher oscillation frequency with lower bias voltage and noise level of TUNNETTs will be superior to those of IMPATTs [41–45]. Transit-time diodes with a tunnel electron emitter are in principle the highestspeed transit-time diodes and are appropriate for the terahertz range oscillatory regime. In contrast to their IMPATT- and BARITT-competitors, they can be completely ballistic devices since the principle of action inherent in them does not require any dissipative processes either in a tunnel barrier or in a transit space [46, 47]. Taking into account physical phenomena, accompanying a hybrid regime between the tunnelling effects and avalanche ionization, is also a difficult problem. That is why different approximations are often used for qualitative and quantitative analysis. A rather high level of material quality and device processing is achieved in fabricating various transit-time diodes made of SiC epitaxial layers. The use of SiC is promising also because the magnitude of electrical breakdown voltage in SiC is higher in comparison with Si and other semiconductor materials. Here we expect an increase in the amplitude of the microwave signal, all other factors being equal. As is well known [32–39, 40–47], the negative dynamic resistance (NDR) effects may be enhanced in transit-time devices if the phase lag of the modulation component of the current, which is in anti-phase with the local electric field, is increased. For this reason, the use of SiC is promising also owing to the fact that the charge mobility in the SiC polytypes is rather small, which should lead to an increase in the phase delay between the current and the alternating electric field in the microwave range and, hence, to an increase in the absolute magnitude of NDR and available microwave power (see, for example, figure 11 [36, 37]). In particular, IMPATT diodes [48, 49], fast resonant tunnelling diodes, and tunnel emitter transistors [50–56] made of silicon carbide and having excellent dc and RF performance have been proposed and realized. These devices are being developed for microwave power amplifier and oscillator applications. The existence of deep donor levels and electron traps in the bandgap of channel is assumed. Highquality heterostructures made of different SiC polytypes were reported in [57]. Figure 11. Maximum available microwave power as a function of the trap concentration level for various BARITTs made of Si, GaAs, GaP and SiC [37]. Si Si p+ n, εs1, νs1 la ld1 SiC SiC n, εs2, νs2 n+ ld2 Figure 12. Structure of double-velocity IMPATT diode. A double-velocity IMPATT diode has been proposed in [58], where the IMPATT diode incorporates a heterojunction made of materials having different scattering-limited velocities in the drift zone of the device. It was shown that for different ratios of the scattering-limited velocities in the two drift zones around the heterojunction, a conversion efficiency of as high as 50% is possible to obtain. On the other hand, the parameters of the material, which determine and limit the power characteristics of diodes, are the following: the critical field Em , where avalanche ionization takes place, saturation velocity of charge carriers and thermal conductivity. The value Em in SiC is about an order of magnitude higher than the breakdown field for Si or GaAs. The saturated electron velocity and the thermal conductivity are higher than those of Si. This means that more heat can be dissipated successfully in SiC. In view of the parameters mentioned, rather small lattice SiC mismatches with AlN, GaN, Si, as well as the abundance of polytypes in SiC, make such a material very promising for manufacturing different heterojunction electronic devices using different bandgap, charge carrier, mobility, etc for microwave and millimeter wave power sources [2, 6, 33, 42–67]. The purpose of the paper [58] was to show how to use different hetero structures (for example, Si/SiC, 3C–SiC/6H–SiC, 3C–SiC/4H–SiC) for fabrication of doublevelocity SiC-based IMPATT diodes. The difference in mobility and velocity can lead to high efficiency and power performance. The structure of the device consists of a uniform avalanche region followed by a drift one and the heterojunction is introduced into this drift region. The structure can be either p+ –n–n–n+ -type, as shown in figure 12, or n+ –p–p–p+ , so the avalanche multiplication takes place in p+ –n or n+ –p junction, while the drift region consists of the n–n or p–p heterojunction. 6361 V V Buniatyan and V M Aroutiounian When l1 = 0, the whole drift layer is made of SiC producing a single velocity heterojunction IMPATT diode. When l2 = 0, the drift region is made of Si only and then the homo-junction IMPATT diode made of Si is (ε1 = ε0 εs1 , ε2 = ε0 εs2 , ε0 is the permittivity of vacuum, εs1 and εs2 , 1 and 2 , νs1 and νs2 are the dielectric permittivity, the widths and saturation velocities of the first and second drift zones, respectively) realized. In all other cases the drift region is constituted by both Si and SiC and the double-velocity heterojunction IMPATT diode is realized. It was shown in [57, 58] that the resistance of the heterojunction Si/SiC IMPATT diode becomes negative in a wider frequency band (f ≈ 36–60 GHz) in comparison with the single velocity Si homo-junction IMPATTs, all other factors being equal. This is due to the higher saturation velocity of carriers in SiC. At a given frequency the magnitude of NDR can be optimized by varying the widths of the first and second drift regions. Moreover, when we have the double-velocity IMPATT diode, the maximum NDR effects are more sensitive to the value of l1 , while the frequency band to that of l2 . To increase the NDR in absolute value, it is more important to control the width of the first drift region where mainly the phase lag of ac current with the local electric field is formed. If l1 = 0 and l2 = 0, the NDR effects weaken when l1 and l2 become larger than 3.0 µm. In all cases, for larger NDR at high frequencies (therefore, for the high output), the requisite length of the first drift region should be larger than that of the second drift region, while at low frequencies it is just the opposite due to the difference in the saturation drift velocities in the two materials. So the performance of the double-velocity heterojunction IMPATTs is better in comparison with the single velocity homo- and heterojunction IMPATTs, all other factors being equal. The possibilities of using SiC in microwave technology are especially attractive, in particular, for manufacture of BARITT diodes [34–39]. Since tunnelling is the fastest phenomenon observed in semiconductors, the quantum well (QW) tunnelling structures have recently drawn a great deal of attention, including BARITTs containing QWs in the injection and drift regions. We report below the achievements in SiC BARITTs. Semiconductor structures with QWs and dots have recently attracted considerable attention (see, for example, [68, 69] and references within). In such a structure the charge transport process suffers qualitative variations connected with the size-quantization energetic spectrum and quantum-mechanical tunnelling of charge carriers through thin potential barriers. The electrons trapping in and escaping from QWs have a remarkable influence on all the generationrecombination processes taking place in such devices. A change in the physical principles of the operation is also opening prospects for the fabrication of new high-frequency (HF) devices [32–35, 67–76]. Experimental investigations and numerical simulations related to the charge transport properties of SiC heterostructures with QWs were not fundamentally considered before [67, 68]. Any information about QWs microwave structures made of SiC is not available in the scientific literature despite the possibilities of manufacturing QWs into SiC structure. The theoretical analysis of the possibility of manufacturing BARITT diodes made of SiC with QWs in the drift region was carried out in [68, 69]. Here the 6362 following were discussed: – the possibility of using 3C–SiC/Si, 6H–SiC/3C– SiC, GaN/6H–SiC, AlN/SiC and related materials for manufacturing injection transit-time devices with QWs; – the performance potential of QW BARITTs in comparison with other transit-time devices. Results of simulations were interpreted in terms of optimum design and maximum frequency of oscillations. The number of QWs M, their thickness LW , and parameter β = τe /τc (where τe is the time of escape from and τc the time of capture by the QW, respectively) were the three parameters used to optimize the device performance. It was taken into account that the primary disadvantages of the Si BARITT diodes are the comparatively small values of the negative dynamic resistance, output power and efficiency [36–39, 68–69]. These drawbacks can in part be removed by various methods. The negative resistance effects can be increased in BARITT structures if the phase lag of the density modulation component of the current, which is in anti-phase with the local electric field, is increased. In some cases this can be achieved by the use of a low-mobility material in order to reduce the velocities of space charge carriers in the source region. Then, after the transit through the central and drain regions, phase lag and generated power are increased. The authors of [68, 69] supposed that the creation of a definite quantity of quantum wells (QWs) in the drift region is promising to optimize the microwave parameters of BARITTs. It was assumed that, analogously to the case considered earlier for usual BARITTs, the capture of the injected charge carriers by QWs decreases the number of the free charge carriers and simultaneously creates immovable space charge which can essentially change the form of the current–voltage characteristic and impedance of the BARITT diodes. By varying the parameters and the number of QWs, it is possible to control both the electron capture time and their escape time from QWs. This allows controlling the ‘effective transit-time’ of charge carriers and increased phase delay (lag) between the alternating current and applied voltage. A n+ –p–n+ BARITT structure, in the p-type drift region of which the presence of QWs for the electrons is assumed (figure 13), was considered. Such a device was termed a barrier injection quantum well transit-time (BARIQWTT) diode. It was assumed that all QWs have the same parameters (the well depth E and the width LW ) and that they are separated by the same barrier layers of thickness LW . As in the case of usual BARITTs, the voltage applied to the structure is assumed to be greater than the voltage UPT , which is necessary for realization of the punch-through condition. It was also assumed that the electric field is not high enough to cause the impact ionization of QWs and the generation of electron–hole pairs in the high field region due to avalanche multiplication processes. So, all QWs are able to trap and keep electrons and the field ionization of QWs is absent. When a weak sinusoidal signal is superimposed on the bias voltage, electrons are periodically injected from the forward biased junction over the barrier into the drift region during the negative half-cycle when the overall voltage is greater than the punch-through voltage UPT . In such a structure a part of the injected electrons is captured by QWs during the time of their drift through the base. As a result, an immovable space charge appears and the concentration of Wide gap semiconductor microwave devices p n Lb + Lb n + EF EF Ec EC LW EF à) n1 n OS EC n QW nK n QW b) EF n QWK EC Figure 13. Schematic diagram of the potential profile of the conduction band in the n+ –p–n+ structure under different applied voltages: (a) U = 0, (b) U
UPT [68, 69]. free electrons decreases. At the same time, in addition to the capture processes, thermionic emission of electrons from QWs occurs. The sequential tunnelling between QWs is scarcely a probable process if the widths of inter-well barriers are greater than 10 nm. The above-mentioned processes can be characterized by the time of capture τc , the time of escape (reemission) τe , and the time of drift τd , respectively. The relation between free and captured electrons in the drift p-region depends on these times and on the level of the injection current. If the injected charge carriers can be captured by QWs and reemitted by them repeatedly during a half-period of the HF signal, charge carriers are delayed during the process of their transport through the base to the drain. This leads to an increase in the phase lag between the alternating current and voltage. According to the general principles of formation of negative dynamic resistance (NDR) in the transit-time structures, such an increase in the lag leads to an increase in the absolute value of NDR. It is assumed that the level of the injection of charge carriers is rather low and the re-distribution of the electric field between QWs can be neglected. The electron capture by QW is usually associated with the process of emission of optical phonons and is characterized by a typical time of the order of ∼1 ps [70–76]. The drift time τd = Lb /vd ∼ 0.1 ps is also rather small if Lb ∼ 10 nm, vd ∼ 107 cm s−1 . Therefore, it can be considered that the rate of the re-emission of charge carriers from QWs is lower than the rates of capture and drift (τe > τc , τd ). In standard BARITTs, where QWs are absent, NDR appears in the range of the transit angles π < θ < 2π with the optimum value θ0P = 1.5π for the maximum value of NDR. If QWs are present, NDR is formed in a more narrow range of transit angles, π/α < θ < 2π/α, where α
1 is a function of the above-mentioned parameters and concentration of QWs in the transit-time region. Note that phase lag effects caused by QWs are similar to the influence of traps described in [36–39], i.e. the presence of QWs leads to an increase in the ‘effective transit-time’ of injected charge carriers. Under otherwise equal conditions, with the increase in the parameter α and the concentration of the QWs, the NDR increases in absolute value, but the frequency band, where NDR takes place, narrows and shifts towards the lower frequency range. Note that the operating frequency range of the BARIQWTTs can be one and more orders of magnitude higher than that of the standard BARITTs. As was mentioned above, TUNNETTs with higher oscillation frequency, lower bias voltage and low noise level have been evaluated as useful devices in the frequency range from 100 to 1000 GHz [33, 40, 41, 43]. GaAs TUNNETT diodes with the p+ –n and p+ –n–n+ structure have been fabricated and pulsed fundamental oscillation frequency of up to 338 GHz (λ = 0.89 mm) has been obtained from p+ –n– n+ diode [11, 41–43]. Simulations of the impedance quality factor, Q, and of the noise in SiC TUNNETTS were carried out in [44]. Calculations of the noise temperature, Ten = (Ms + MT )T0 , were made as in [44] using the assumption that the most probable sources of the noise in such TUNNETTs are shot (fluctuation, Ms ) and diffusion (thermal, MT ) noises. Here T0 is 290 K. The use of 4H–SiC is promising for the manufacture of high-frequency TUNNETTs due to the fact that after tunnelling the charge carriers are moving at their saturation velocities. The use of SiC makes it possible, all other factors being equal, to increase the amplitude of the microwave signal and, consequently, the efficiency and power performance. With the increase in the injection current density, effective small-signal tunnelling conductivity in the tunnelling plane and charge velocity, NDR increases in absolute value and the frequency range where NDR takes place becomes wider and is shifted towards higher frequencies. The noise measure (temperature) decreases monotonically and in the region, where NDR take place, remains approximately constant, the noise inevitably increases roughly. After numerical calculations the conclusion is drawn that the performance potential of TUNNETTs is better in the very high-frequency range than that of the other transit-time devices. The use of silicon carbide for the manufacture of TUNNETTs promises better microwave parameters over the frequency range 100– 500 GHz. Hence, the following conclusions can be made. Significant progress is obvious in the development of SiC technology, but important questions remain connected with the effects of post-oxidation anneals in various ambients, the effects of surface orientation and polytypes, and the long-term reliability of oxides on SiC. For example, insulators capable of reliable high-temperature operation with high breakdown field strength and lower interface trap densities are needed for further development. It is possible to conclude on the basis of thermal analysis that, due to the much smaller value of off-state power loss for SiC devices as compared with Si devices, much higher on-state current densities (power) can be achieved in SiC devices for given junction temperatures and packaging. Due to the high thermal conductivity and electric breakdown field strength of SiC, it is possible to achieve integration of SiC devices at higher packaging densities. However, this would require considerable improvements in SiC device processing technology. Devices, such as high-frequency transit-time structures and MESFETs, which do not rely on large area or high-quality dielectric may be the first SiC electronic (excluding optical) devices to hit the commercial market. 2.1. SiC-based diodes The principal advantages of the SiC Schottky rectifiers and p-i-n diodes are their superior switching and transient 6363 V V Buniatyan and V M Aroutiounian characteristics with high turn-off speed and the absence of a large reverse recovery current flow [11, 12, 42, 49, 77–88]. There are two basic classes of rectifiers, namely Schottky and p–n junction. The former have lower on-state voltages and higher switching speeds, while the latter have higher breakdown voltage and lower reverse leakage current [86]. Being a majority-carrier transport device, the SiC Schottky rectifier operates without high level minority carrier injection. Compared with the Si Schottky rectifier, the SiC Schottky rectifier is expected to demonstrate higher currenthandling capability and low forward-voltage drop at high breakdown voltages. Unlike the Si Schottky rectifier having a breakdown voltage of less than 100 V, an ideal 5000 V 6H–SiC rectifier can deliver 100 A cm−2 with a forward-voltage drop of less than 3.85 V [81]. A 5000 V diode made of SiC requires only a 40 µm thick drift layer instead of almost 500 µm for Si diode. The forward current-voltage characteristics of 4H– SiC p+ –n diodes with a 5.5 kV voltage-blocking capacity with the hole lifetime in the base about 1 ns were observed in [78]. The best n-type 4H–SiC diode with a 10 kV blocking voltage (100 µm base width) had the breaking hole lifetime τp = 3.7 µs at room temperature [82]. As the thermal stability of ohmic and rectifying contacts in SiC-based devices is extremely important, an additional barrier height optimization was needed. As the thermal stability of Ohmic and rectifying contacts in SiC-based devices is extremely important, and as it has been shown that the Fermi-level pinning [49] in SiC Schottky contacts is absent, an additional barrier height optimization is needed. Thus, the SB height depends on metal work function. It has been shown [77] that the SB height can be reduced (0.19–0.25 eV on 4H–SiC and 0.15– 0.17 eV on 6H–SiC for n-type Schottky contacts, and 0.02– 0.05 eV for 4H-p-SiC and 0.1–0.13 eV on 6H contacts) using Au nano-particles. It has also been stated [49, 79, 80] that the annealing regime and atmosphere mainly affect the reverse biased characteristics (the reverse leakage current is reduced) as a result of the decreasing number of surface states. The main obstacle to the implementation of high-power SiC p–n junction diodes is forward-voltage drop degradation (voltage drift) during long-term operation [84]. This phenomenon was observed in p-i-n diodes fabricated by both ion implantation and epitaxial growth techniques. Intensive investigations by a number of groups have demonstrated that the voltage drift is caused by the generation of crystallographic defects (specifically stacking faults) in the device active region (see [84] and references therein). Results of the development of microwave 4H–SiC p-i-n diodes with optimized device structure as well as the performance of X-band switches were reported in [85]. The diodes with mesa-structure diameters between 80 and 150 µm had a blocking voltage of 1100 V, a 100 mA differential resistance of 1–2  and a capacitance below 0.5 pF at a punch-through voltage of 100 V, and the carrier effective lifetime 15–27 ns. The switches exhibited an insertion loss of as low as 0.7 dB, isolation up to 25 dB and were able to handle microwave power of up to 2.2 kW in isolation mode and up to 0.4 kW in insertion mode [85]. 3. GaN- and AlGaN/SiC-based devices Group-III nitride semiconductors and heterostructures have been recognized as belonging to the most promising 6364 materials for high-temperature high-power high-electron mobility transistors (HEMTs) [4, 5, 89–98], optoelectronic devices in the short wavelength region (photodetectors and LEDs) [99–107], monolithic microwave and optoelectronic integrated circuits [108–117], heterojunction FETs, MISHFETs, MOSFETs [118–150], microwave, heterojunction and switched diodes (heterodiodes) [130–137], lasers [109, 138–140], as well as other devices, which will have a great impact on the future world [140]. The superior physical and chemical stability of the nitride semiconductors will enable them to operate in harsh environments. To produce such novel devices, it is essential to grow highquality nitride single crystals and to control their electrical conductivity. For example, it was quite difficult to grow high-quality epitaxial GaN films with a special crack-free surface. The recent progress in growth technology of extremely high-quality GaN single crystals with such a surface, the discovery of p-type GaN and the ability to fabricate a p–n junction light emitting diode [135] have led to such developments as high-performance blue and green LEDs, violet laser diodes, ultraviolet (UV) photodetectors and field effect transistors. GaN and related III–N materials are typically grown on lattice mismatched substrates such as Si, sapphire and SiC. A consequence of this mismatch is that III–N epitaxial layers contain a high density of threading dislocations causing degradation of laser diodes, which operate at a high current density (2–4 kA cm−2 ). They may pose an obstacle to the high-performance of photodetectors and transistors [135, 136]). Various technologies have been developed for the reduction of the dislocation density. These approaches include lateral epitaxial overgrowth [137, 138] and cantilever epitaxy [149], and significantly reduced dislocation densities in the range 106 –107 cm−2 have been achieved. Theory [140] and experiment [141] indicate that a further reduction in defect density is possible if the lateral overgrowth approach is extended to the nanoscale. For growing nitride films the pulsed laser deposition method has been reported in [142]. It was found that crystallinity, surface morphology, and optical transmittance of films may be improved with an increase in the deposition temperature. The surface morphology of GaN may be improved also by using a low-temperature-deposition buffer-layer technology (for example, AlN) [135]. For the formation of n- and p-type conductivity, the usual ion implantation into the crystalline material can result not only in damage of the crystal structure but also in a multitude of point defects and cluster defects. These defects usually behave as luminescence quenching centres. These centres, namely electron or hole traps, correspond to the same deep levels in the forbidden gap [143]. Among photoluminescence bands related to deep centres, research has also focused on the origin of the so-called yellow luminescence band located at ∼2.2 eV and transition metal-impurity-related bands [143]. Further progress depends on the detection characterization and eventual reduction of defects near the two-dimensional (2D) electron gas interface in HEMTs and FETs [144]. Trapping centres near the interface not only reduce the 2D gas carrier densities but they also play a role in HEMT device current hysteresis. GaN device technology is by now strongly polarized between photonic and electronic applications. Wide gap semiconductor microwave devices Minority carrier diffusion length is the crucial material parameter also for the base of heterojunction bipolar transistors, affecting the device design and overall performance. The other important material parameters are the anisotropy of the minority carrier diffusion length, dislocation density and doping level [145]. For heterojunction BTs made of nitride semiconductors, improvements in p-type doping are needed. At present, due to a low hole concentration in the base at room temperature, the access resistance is high. A second concern is in achieving sufficiently long minority carrier lifetimes (and diffusion length) in the base, i.e. realizing high current gain, which at the present time are only around 10. On the other hand, the forward voltage of the chip is significantly affected by the p-contact properties, so this electrical characteristic is of key importance in customer specification. The contact resistance leads to an additional contribution to the total voltage drop of the device under operation. Thermionic and field emission of holes in the valence band from the metal into the p-GaN are the two main current transport mechanisms for Schottky contacts. From the viewpoint of good p-contact technology, a low Schottky barrier is recommended which increases the thermionic emission, as well as a high net acceptor concentration near the interface, which lowers the thickness of the depletion region and so increases the field emission. From the theory of barrier height [11] B is determined by the metal work function m , bandgap Eg and electron affinity X of the p-GaN: B = m − (X + Eg ). (4) For GaN (X ≈ 4.1 eV, Eg ≈ 3.4 eV) this means that a metal is not available, since one still obtains positive barriers even for materials with the highest work function (5–6 eV). Apart from this fact, there are always interface states in the bandgap that can strongly influence the Fermi level near the surface and therefore also the determined barrier height. Investigations carried out in [135] show that the quality of p-contact and therefore the value of the forward voltage are very sensitive during the chemical preparation of the epitaxial layers before metallization, since, on the one hand, a chemically well defined surface is necessary for good contacts and, on the other hand, extremely aggressive chemicals such as acids attack the p-GaN material resulting in an increase in the voltage. Furthermore, it is possible to check the contact resistance at any step of the chip fabrication process. Most of the works which are carried out in the field of the GaN/AlGaN electronic and optoelectronic devices usually addressed the hexagonal phase of GaN, which is thermodynamically stable, but in which very high spontaneous and piezoelectric fields are present [102]. Such internal fields strongly influence optical and electronic properties of heterostructures and device performance for either light-emitting device or high-mobility transistors. Therefore, the experimental determination of such internal electric fields is unavoidable for the device design, which has been done in [105, 106, 125] in the Franz– Keldysh regime for the GaN/AlGaN quantum wells. It is found that the field strength is ∼850 kV cm−1 in a 4.2 nm GaN quantum well. Photoreflectance studies were carried out on AlGaN/GaN heterostructures confining high-mobility polarization-induced two-dimensional electron gases [106]. By analysing the Franz–Keldysh oscillations for samples with Table 9. A summary of the compared diodes, which are both taken from our own data and from other groups [135]. Sample Buffer MOCVD-1 120 A A1N MBE-1 200 A LT-GaN MBE-2 100 A LT-GaN HVPE-1 None HVPE-2 None None None None None None SiC layer Polytypes EA Ec p-epi, 1 × 1016 cm−3 p-epi, 1 × 1016 cm−3 p-epi, 1.4 × 10 (cm3 ) p-epi, 3 × 1015 cm3 p-epi, 3 × 1015 cm−3 p-epi, 1.4 × 1016 cm−3 p-bulk p-bulk p-bulk n-bulk 6H — −0.3 6H 1.7 — 4H 1.8 −0.9 4H 2.1 −1.1 4H 2.1 −1.0 4H 2.8 0.1 6H 6H 6H 6H — — — — −0.1 −0.83 −0.7 −1.1 Note: The estimated conduction band discontinuity is also included. (Al) Ga and N-face polarity, values of the surface electric field up to 380 kV cm−1 at room temperature were obtained. Taking into account spontaneous and piezoelectric polarization, the density of changed surface states and the bare surface potential are estimated. The results prove the presence of donor- and acceptor-like surface states with (Al) Ga- and N-face polarity, respectively. A change of the electric field was observed upon exposure to a polar liquid demonstrating the applicability of these structures for chemical sensors. 3.1. Nitride based diodes (contacts) It was mentioned above that SiC is a semiconductor material used for high-power and high-temperature applications, but it also plays an important role as a substrate material for GaN/AlGaN devices [136]. The MOSFETs and insulator gate bipolar transistors based on SiC are the dominating types for high-power and high-voltage devices of silicon. For SiC, however, the oxide/SiC interface has low channel mobility, which has been a severe obstacle in fabricating these SiC devices. The use of heterojunctions for the base– emitter (in BJT) junction improves the efficiency of the emitter injection substantially, even for a very high base doping. For SiC, a possible heterojunction is the GaN/SiC n–p junction [134]. As the heterojunction can improve the performance considerably for LEDs, laser diodes, as well as HBTs and HFETs, extensive investigations of GaN/SiC and AlGaN/SiC heterodiodes [128–137], grown with different growth techniques, are carried out (table 9 [135]). The ideality factors in such diodes for different temperatures are summarized in table 10 [134]. A low turn-off voltage is typical for the GaN/SiC heterojunction [134]. It was compared and included in a model for a recombination process assisted by tunnelling, which was proposed for explanation of the low turn-on voltage [134,135]. The model predicts that the conduction band offset is in part responsible for the high rate of recombination processes. The band offset has been calculated from the Schottky barrier measurements, resulting in a type II band alignment with 6365 V V Buniatyan and V M Aroutiounian Table 10. Ideality factors from the I –V measurements at different temperatures [134]. Sample Diameter (µm) Temperature (◦ C) Ideality factor, η MBE MBE MBE MBE MBE MBE MBE 50 50 100 50 100 50 100 25 100 100 150 150 200 200 1.28 1.24 1.22 1.23 1.20 1.20 1.19 Figure 14. The drawing shows the band diagram of a GaN/SiC heterojunction, where the injection process and tunnelling assisted recombination process are indicated by arrows. This band alignment is of type II, and will result in increased carrier concentrations on both sides of the junctions [134]. Figure 15. Comparison of the results of measurements and simulations at 25 and 300 ◦ C for the HVPE-1 diode and the MBE-2 diode. The band discontinuity, Ec , was 1 and 0.8 eV for the HVPE-1 diode and the MBE-2 diode, respectively. The recombination velocity was set to 3 × 107 cm s−1 for all simulations [134]. a conduction band offset in the range 0.6–0.9 eV. Since the conduction band of GaN is lower than that of SiC (figure 14), a high electron concentration can be located very close to the junction. Similarly, the valence band of SiC is above the valence band of GaN, resulting in a higher concentration of holes close to the junction. These carrier concentrations increase when the junction is forward biased (figure 15) [134]. In table 11 the calculations for all the investigated diodes are summarized. The data extracted from measurements were used for the calculation of Ec (or EA ) assuming that EA (or Ec ) is 6366 known from the measurements. The calculated ideality factor was close to unity for all diodes. As can be seen in table 11, the calculations agree well for both HVPE diodes in contrast to the NBE diodes. To further investigate the band alignment in the GaN/SiC heterojunction, several measurements on SB height have been performed and the results collected in table 12; these values have been used to estimate the band offset on the basis of the midgap theory: Ev,AB = BP,A − BP,B , EC,AB = Bn,A − Bn,B , (5) where Bn and BP are SBs for n-type and p-type materials, respectively. The subscripts A and B refer to the two materials constituting the heterojunction. The midgap theory assumes that there exists an energy level that determines the thickness of the dipole layer between the materials. If the affinity rule is valid within 0.5 eV and the susceptibility of the materials are close to 10, the midgap theory predicts the band offset within 0.05 eV. However, this theory also predicts that the SB height, when a Schottky contact is made of a certain semiconductor, is independent of the metal. There are several examples showing that this is not the case, although the work function also fails to model the SB. This discrepancy has been attributed to nonideal effects such as Fermi-level pinning [147]. The Schottky barrier data collected from literature were, when possible, selected in agreement with the Schottky–Mott limit, i.e. that the sum of the p-type and n-type barriers should be equal to the bandgap. Such references were found for both 6H–SiC and 4H–SiC [129, 134, 135], but none was found for GaN or AlGaN because of poor p-type Schottky contacts. As a result, only the n-type contacts may be used for GaN and AlGaN, and the p-type barrier should be calculated in the Schottky–Mott limit. The collected SB heights are summarized in table 11. The results of the band offset calculations are given in table 12. Both C–V measurements and the SB investigation indicate that the conduction band offset between 4H–SiC and GaN is in the range −0.7–1.0 eV, the band offset is of type II (a negative value of Ec ). This higher negative band offset is the origin of the low turn-on voltage of the GaN/SiC heterodiodes. A modified expression for the SB height that takes into account the effect of total polarization and the 2DEG in GaN at the GaN/AlGaN heterointerface through the incorporation of the SB lowering is investigated [129]. Updated SB heights for HEMTs and their bias dependence are studied in [129] for two values of the barrier height, φb = 1.33 eV and φb = 1.19 eV. Figures 16 and 17 show the simulated I –V characteristics, transfer characteristics and transconductance for these two barrier heights. Due to a high contact resistance of p-GaN originating from the low doping density of p-GaN, many studies have been made to improve the poor electrical performance. In [133], instead of partially transparent multilayer metals (such as Ni/Au), an indium oxide contact to p-CaN is used and has been investigated. Indium tin oxide is well known as a transparent conducting material with the resistivity of about 10−4  cm2 . Therefore, indium tin oxide can be used to form a transparent ohmic contact on p-GaN. The optimization of HEMTs, LEDs and other devices made of GaN/SiC and AlGaN/SiC is still in its infancy and is yet to account for the effects of spontaneous and piezoelectric Wide gap semiconductor microwave devices Table 11. A summary of the calculations using the interface recombination model [134]. Measured values −2 Calculated values Sample JSat . (A cm ) η EA (eV) Ec (eV) EA (eV) Ec (eV) Sj (cm s−1 ) MOCVD-1 (100 ◦ C) MBE-1 (100 ◦ C) MBE-2 HVPE-1 HVPE-2 SiC-1 — 3 × 10−19 7 × 10−27 8 × 10−26 1 × 10−27 9 × 10−25 — 1.3 1.3 1.1 1.1 2.0 — 1.7 1.8 2.1 2.1 1.8 −0.3 — −0.8 −1.1 −1.0 0.1 — 2.6 2.4 2.1 2.2 — — −1 −1.4 −1.1 −1.1 — — 1 × 102 4 × 102 5 × 108 3 × 106 — Table 12. A summary of Schottky barrier heights measured on 4H–SiC, 6H–SiC, GaN and AlGaN, where the bowing parameter for the AlGaN bandgap calculation was 1.3 [134]. Wide Schottky bandgap Bandgap Schottky Extraction Dopant barrier semiconductor (eV) metal method type (eV) 4H–SiC 3.26 Ni Pd Au 6H–SiC 2.98 Pd Au Ni GaN 3.39 Au Ni Pd AlGaN A" = 0.11 3.57 Ni A = 0.15 3.65 Ni J C = 0.17 3.68 Ni IV CV IPE IV CV BEEM IV CV IPE IV CV XPS IV CV XPS IV XPS XPS IV CV XPS XPS CV CV AE CV AE IV CV PL CV AE IV CV PL IV CV PL IV CV PL n n n p p n n n n p p n n n p n n p n n n p n n n n n n n n n n n n n n n n n n n 1.62 1.75 1.69 1.31 1.48 1.54 1.73 1.85 1.81 1.35 1.42 1.11 1.11 1.26 1.61 1.37 1.40 1.45 1.29 1.29 1.24 1.56 0.98 0.98 0.86 1.13 0.99 0.84 1.00 0.97 1.07 0.90 0.94 1.24 1.25 1.04 1.26 1.29 1.11 1.36 1.33 Note: CV stands for capacitance–voltage, IV for current–voltage, PL for photoluminescence, AE for Arrhenius plots, BEEM for ballistic-electron emission microscopy, IPE for internal photoemission spectroscopy and XPS for x-ray photoelectron spectroscopy. Figure 16. I –V characteristics with barrier height as a parameter. For gate biases of Vg = 0 V and −1 V and the data shown, the sample was illuminated before measurement [129]. Figure 17. Variation in transfer characteristics and transconductance with barrier height as a parameter Vd = 3.0 V [129]. polarization on the device performance parameters. One such parameter is the SB height and its importance stems from the fact that it relates to breakdown voltage, leakage current, charge control, etc. The SB for HEMTs, taking into account the effect of total polarization and the 2DEGconcentration in GaN at the GaN/AlGaN heterointerface through the incorporation of the SB lowering, is investigated in [134]. It is stated that the variation of spontaneous polarization due to the finite pyroelectric coefficient of AlGaN and the piezoelectric component due to thermal expansion of the lattice constant, may suggest that the SB height will be temperature 6367 V V Buniatyan and V M Aroutiounian Table 13. Calculated band discontinuities from midgap model compared with different values found in the literature [134]. Interface GaN–4H–SiC GaN–6H–SiC Gax Al1−x N–4H–SiC 4H–SiC–6H–SiC Midgap Ev Au (eV) Ec Midgap Ev Ni (eV) Ec Midgap Pd (eV) Ev Ec Note (0.99) (0.86) — — — −0.07 −0.86 −0.45 — — — 0.41 (0.83) (0.70) (0.85) (0.88) (0.84) −0.17 −0.70 −0.29 −0.54 −0.49 −0.42 0.41 (0.69) (0.59) — — — (−0.10) −0.56 −0.18 — — — 0.38 Calculated p-GaN Calculated p-GaN x = 0.11, calculated p x = 0.15, calculated p x = 0.17, calculated p Calculated p-4H–SiC Note: The value in parentheses has been extracted using the n-type barrier subtracted from the bandgap. A minus sign indicates that the material first in order has a band edge lower in energy than the second one. Figure 18. I –V diagrams for all types of diodes at 100 ◦ C. Al0.3 Ga0.7 N is still shifted towards the 4H–SiC sample [136]. dependent in GaN-based HFETs and MESFETs. However, the calculated values are practically temperature independent. It is also stated [134, 135] that 2DEG depends on the applied gate bias, which in turn affects the barrier height. Figures 16 and 17 show the simulated I –V characteristics, transfer characteristics and transconductance for the two (AlGaN) barrier heights (B1 = 1.33 eV, B2 = 1.19 eV). The electrical properties of GaN and AlGaN heterodiodes have been investigated with varying Al compositions in [136]. The characteristics (both C–V and I –V ) were similar for all diodes except for the Al0.3 Ga0.7 N diode, which had a clear change in I –V (figure 17, table 13 [136]). The C–V measurements showed that the Ec is almost independent of the Al compositions, which can indicate the presence of a strong interface charge at the junction. A model for a tunnelling assisted recombination process was analysed for the GaN diode (figures 17 and 18), since this diode was the only one showing a consistent built-in-potential (activation energy) from I –V measurements (table 14). The proposed tunnelling assisted recombination can be due to a combination of the lattice mismatch and the GaN/SiC heterojunction band alignment, where Ec (GaN) < Ec (4H–SiC) or of the type II. This band alignment type produces higher carrier concentrations close to the interface at forward bias, and the lattice mismatch can introduce high concentrations of recombination centres. The band alignment of AlN has Ec (AlN) > Ec (4H–SiC) and the lattice match for AlN is only 1%, so these properties should gradually improve with an increase in the Al composition for Alx Ga1−x N. To improve the ohmic contact characteristics for the p-GaN optoelectronic devices by means of Mg doping, the 6368 original SB height as high as 2.4 ± 00.2 eV is obtained [137]. The contacts showed significant memory effects in I –V depletion-layer capacitance with a time constant as long as 8.3 × 103 min at room temperature, resulting from carrier capture and emission from a high density of deep level defects. Reliable, low-resistance ohmic contacts to AlGaN/GaN HEMT structures are essential for efficient operation. Ohmic contact processing is still a challenging area in the AlGaN/GaN device technology. For this reason, two metal stacks, Ti/Al/Ti/Au and Ti/Al/Mo/Au, were evaluated for making ohmic contacts to AlGaN/GaN structures [128]. Both metal stacks were optimized with respect to metal layer thickness and annealing time. Optimum contact resistance of 0.54  mm was obtained for the Ti/Al/Mo/Au metal stack on AlGaN/GaN/sapphire with Al thickness of 90 nm, 1 : 6 Ti : Al ratio and Mo thickness of 40 nm annealed at 850 ◦ C for 1 min. The same optimized stack applied to a structure on an SiC substrate demonstrated higher than 0.8  mm contact resistance. Optimized Ti/Al/Ti/Au (15 nm/80 nm/45 nm/80 nm) metal stack on a sapphire substrate showed 0.48  mm contact resistance on being annealed at 850 ◦ C for 2 min. Optimized sheet resistance for the Ti/Al/Ti/Au structure on the sapphire substrate was 420  sq−1 and for the structure on the SiC substrate 300  sq−1 . Ti/Al/Mo/Au stack on sapphire substrate had a slightly higher contact resistance compared with the Ti/Al/Ti/Au stack, but much better edge acuity and surface roughness. There was no metal spilling or creeping noticed for the Ti/Al/Mo/Au stack, in contrast to the Ti/al/Ti/Au metal stack. Ohmic metal was etched to expose AlGaN/metal interface. Conductive defects were found on the surface, while the rest of the AlGaN surface was found non-conductive as well as unalloyed regions. These conductive defects (inclusions) are believed to be alloyed screw dislocations serving as conductive paths to the 2DEG. These experiments showed that the AlGaN layer was not alloyed efficiently during ohmic contact formation. Only discrete conductive paths were formed, which result in poor reliability, fast degradation under stress conditions and poor contact resistance to the 2DEG. There is a need for developing alternative ohmic contacts to improve contact resistance and reliability of these contacts to the AlGaN/GaN structure. 3.2. GaN-based transistors GaN/AlGaN high-electron-mobility transistors (HEMTs) can offer even higher power performance due to the higher carrier Wide gap semiconductor microwave devices Table 14. A summary of the extracted parameters for the measured diodes from I –V measurements, where the errors are estimated with standard deviation [136]. Cathode material Built in voltage (V) Saturation current at 25 ◦ C (A cm−2 ) Ideality factor Series resistance () Sheet resistance (/ ) GaN Al0.1 Ga Al0.15 Ga0.85 N Al0.3 Ga0.7 N Al0.5 Ga0.5 N 4H–SiC 2.08 1.15 1.26 0.94 0.95 1.76 1.1 × 10−27 7.5 × 10−19 6.9 × 10−20 4.9 × 10−16 1.1 × 10−17 8.9 × 10−25 1.05 ± 0.04 1.53 ± 0.10 1.47 ± 0.11 2.63 ± 0.24 1.81 ± 0.09 1.96 ± 0.08 4.2 × 105 2.0 × 105 4.9 × 104 8.8 × 103 4.2 × 105 2.8 × 103 90 1250 1950 1750 1440 310 sheet density and saturation velocity of 2DEG compared with SiC, and a record power density of 10 W mm−1 has been demonstrated from 50 µm or 150 µm gate periphery devices [3, 4]. With the increased device gate periphery, the power performance will be degraded because of the significant device self-heating and trapping effects. GaN/AlGaN HEMTs have demonstrated more than 6.7 W mm−1 from a 400 µm device tested at 10 GHz. X-ray topography and optical defect mapping have revealed the improvements of micro-pipe density in SiC semi-insulating substrates and lowering of the defect density in AlGaN and GaN epitaxial films. Surface passivation on the GaN/AlGaN HEMTs showed almost full recovery of DC drain current dispersion from surface states. The GaN/AlGaN HEMTs process starts with the growth of epitaxial GaN and AlGaN layers on 4H–SiC semi-insulating substrates. The fabrication process includes mesa isolation, ohmic metal evaporation and annealing, overlay metals, e-beam patterned T-gate with 0.2–0.3 µm footprint, SiNx surface passivation, air bridge crossovers and backside bias (not yet applied). Recently the AlGaN/GaN FETs have been pursued for application in high-power and high-frequency electronic circuits. The GaN HEMTs can operate up to 12 W mm−1 at 10 GHz [129] with fT = 10 GHz and fmax = 155 GHz [131]. Power density as high as 32.2 W mm−1 at 4 GHz from an AlGaN/AlN/GaN HEMT on a SiC substrate has also been reported [130]. The superior thermal properties of GaN have resulted in the realization of heterojunction FETs operating up to 750 ◦ C [142]. Dramatic improvements in the performance of molecular beam epitaxy grown HEMTs have been achieved. The maximum HEMTs have been grown using MOCVD techniques. Further progress depends on the detection, characterization and eventual reduction of defects near the two-dimensional electron gas (2DEG) interface. Traps near the interface not only reduce the 2-FEG carrier densities but they also play a role in HEMT device current hysteresis [144]. Power densities as high as 30 W mm−1 at 2 GHz and 10 W m−1 at 8 GHz have been demonstrated [130, 131]. Reliable, low-resistance ohmic contacts to the AlGaN/GaN HEMT structure are essential for efficient operation. Ohmic contact processing is still a challenging area in the AlGaN/GaN device technology [118, 128–132]. III–nitride HFETs are promising devices for high-power energy converters. Compared with SiC, FETs and GaN HFETs have lower specific on-resistance due to the highdensity 2DEG, i.e. above 1013 cm−2 , and high mobility, i.e. above 1500 cm2 V−1 s −1 . The (RON × A)–VBR dependences Figure 19. Comparison of VBR –RON of the AlGaN/GaN HFET devices [120]. for AlGaN/GaN HEMTs (where A is the area and VBR is the breakdown voltage), SiC and Si devices are compared in figure 19 [120]. The contribution of specific contact resistance is significant for devices with VBR below 1000 V, whereas for devices with high VBR the channel resistance dominates. Thus, the challenge in achieving the lowest RON values in the kV-range III–N HFETs is to minimize the breakdown voltage at minimal electrode spacing. Investigations of the effects of ohmic contacts on the leakage in GaN transistors have shown that the increased buffer leakage arises due to the nature of the alloyed ohmic contacts and can be minimized. They can be screened by Si doping or by the two-dimensional electron gas [118]. A novel approach in the high-performance enhancement mode (E-mode) of the AlGaN/GaN HEMTs has been reported [89]. The fabrication technique was based on fluoride-based plasma treatment of the gate region and post-gate rapid thermal annealing with an annealing temperature lower than 500 ◦ C. A zero transconductance (gm ) at Vg = 0 and the off-state drain leakage current of 28 µA mm−1 at Vds = 6 V were obtained. The fabricated E-mode AlGaN/GaN HEMTs with a 1 µm long gate exhibit a maximum drain current density of 310 mA mm−1 , a peak gm of 148 mS mm−1 , a current gain cut off frequency fT of 10.1 GHz and the maximum oscillation frequency fmax = 34.3 GHz. From the point of view of applications, E-mode HEMTs have many advantages for applications as high-frequency power amplifiers and low noise amplifiers [89–99], E-mode allows eliminating negativepolarity voltage supply and therefore a reduction in the circuit complexity and cost. A planar-fabrication technology for 6369 V V Buniatyan and V M Aroutiounian (a) (a) (b) Figure 20. Id –Vd characteristics at high bias from 0.25 µm gate length devices [116]. integrating enhancement/depletion (E/D)-mode AlGaN/GaN HEMTs was used. Investigations [93] of their characteristics have shown that the E/D-mode exhibits an output swing 2.85 V, with the logic-low and logic-high noise margins at 0.34 V and 1.47 V, respectively. A GaN/ultrathin InGaN/GaN heterojunction has been used to provide a back-barrier to the electrons in HEMTs [95]. The polarization-induced electric fields in the InGaN layer raise the conduction band in the GaN buffer with respect to the GaN channel, increasing the confinement of the 2DEG under high electric field condition. These devices showed excellent high-frequency performance, with a current gain cut-off frequency fT ≈ 153 GHz and power gain cut-off frequency fmax ≈ 198 GHz for a gate length of 100 nm. At different biases, a record value of fmax = 230 GHz was obtained. Investigations of the temperature and gate length effects on the dc performance of the AlGaN/GaN HEMT carried out in [116] show that the defect density in the structure grown on the AlN/SiC template is significantly lower than in the structure grown on sapphire. Reverse breakdown voltages of above 40 V were obtained for a 0.25 µm gate length device on both types of substrate (figure 20). Both types of HEMTs showed similar trends of drain current and transconductance with increasing temperature. Extrinsic transconductance of ∼200 mS mm−1 for HEMTs on sapphire and ∼125 mS mm−1 for the device on AlN/SiC were achieved (figure 21). In order to improve the high-frequency characteristics of HEMTs two techniques have been developed in [115, 116]. One is a high-quality, high-Al-composition AlGaN barrier layer grown by PAMBE, the other is SiN passivation by catalytic chemical vapor deposition (Cat-CVD) which made it possible to fabricate thin barrier HFET structures with higher ns 6370 (b) Figure 21. Transfer characteristics for 0.25 µm length devices [116]. and lower channel-sheet resistance. With the use of these techniques, NFETs with ns > 2.5 × 1013 cm−2 for dAlCaN < 10 nm have been obtained in [115, 118]. NFETs with Lg = 0.06–0.2 µm have a maximum drain current density of 1.17–1.24 A mm−1 at a gate bias of +1 V and a peak extrinsic transconductance of 305–417 mS mm−1 . The current gain cut-off frequency (fT ) was 163 GHz, which is the highest value among those reported for GaN HFETs. The maximum oscillation frequency (fmax ) was also high, and its value was derived to be 163–192 GHz. The E-mode Si3 N4 /AlGaN/GaN MISHFET with a 1 µm gate length and a 15 nm Si3 N4 layer inserted under the metal gate to provide additional isolation between the gate Schottky contact and the AlGaN surface, which can lead to reduced gate leakage current and higher gate-turn-on voltage has been developed in [124]. The forward turn-on gate bias of the MISHFETs was as large as 7 V, at which a maximum current density of 420 mA mm−1 was obtained. The small-signal RF measurements showed that cut-off frequencies of the current gain and the power gain were 13.3 GHz and 23.3 GHz, respectively. The experiments indicated that the Si3 N4 insulator offers excellent insulation between gate metal and semiconductor. As the gate leakage current was reduced by employing the MIS structure and the breakdown voltage characteristics were improved, MIS-HEMT have been investigated in [122] with the use of SiO2 , SiN and TiO2 as insulators. The excellent tradeoff Wide gap semiconductor microwave devices (a) (b) (c) (d) Figure 22. Breakdown voltage Vb as a function of Lgd length: (a) MES-HEMT; (b) SiO2 MIS-HEMT; (c) SiN MIS-HEMT; (d) TiO2 MES-HEMT [122]. characteristics between specific on-resistance and breakdown voltage were found to be on a par with the 4H–SiC limit (figures 21 and 22). The breakdown voltage of the MIS-HEMTs increases nonlinearly with the increase in gate–drain length Lgd . The TiO2 insulator exhibited the highest breakdown voltage of 2 kV with the on-resistance of 15.6 m  cm2 for Lgd = 28 µm. On the other hand, the breakdown voltage and on-resistance for SiN MIS-HEMT were found to be 1.7 kV and 6.9 m  cm2 , respectively (figure 23). The main attention in recent years has been devoted to SiC MOSFETs and GaN HEMTs, but the development of high-performance SiC MOSFETs has been hampered due to poor MOS channel mobility and reliability. The investigations carried out in [126] for n-channel GaN MOSFETs on both p and n-GaN epilayer on sapphire substrates showed good dc characteristics (figure 24) with a maximum field-effect mobility of 167 cm2 V−1 s−1 , the best reported to date. Khan et al [127] were the first to fabricate GaN MODFETs. Using n-GaN-Al0.15 Ga0.85 N heterojunction MODFETs, they demonstrated that devices with 4 µm channel length had a transconductance of 28 mS mm−1 at 300 K and 46 mS mm−1 at 77 K. The performance of shortchannel MODFETs is impressive. Wu et al [133] have fabricated 0.2 µm gate length GaN MODFETs, with a sheet carrier density of about 8 × 1012 cm−2 and a mobility of 1200 cm2 V−1 s−1 , which had fT of 50 GHz. These Figure 23. Tradeoff characteristics between specific on-resistance and breakdown voltage [122]. devices have very high saturation currents (800 mA mm−1 ) and transconductance (240 mS mm−1 ), plus breakdown voltages exceeding 80 V per µm of gate–drain spacing. Remarkable improvements in the device performance achieved in the last five years demonstrate a rapid pace of development of III–V FETs for high-power and microwave applications. Since power MODFETs experience high VDS during operation, the effects of high electric fields on their 6371 V V Buniatyan and V M Aroutiounian (a) (a) (b) (b) (c) Figure 24. (a) Output characteristics of the GaN MOSFETs on p epilayer; (b) output characteristics of the GaN MOSFETs on n-epilayer; (c) sub-threshold slope for the GaN MOSFETs on p and n-epilayers [126]. reliability should be evaluated. Several papers have addressed reliability issues in GaNAs–AlGaAs and GaAs–InGaAs MODFETs [134–136]. Hot carrier effects, impact ionization and high-reverse Schottky diode current stress (off-state breakdown) have been considered in these papers. However, none of the papers gives detailed mechanisms of degradation invoking processes in GaAs MODFETs which are known to exist under high-field stress. During high-reverse current stress, electron trapping dominates for the first 50–60 s and then hole trapping and trap creation begin to manifest. The degradation processes bring about a positive shift of Vt , degrade ID and g and increase reverse leakage (figure 25). The high breakdown field of III–nitride materials should help to reduce the downscaling limits, improving the frequency response of transit-time devices. The saturation velocity for GaN is also higher than that of GaAs, and the optical phonon energies are about three times higher, which can also benefit transient transport. Despite the GaN-based device technology being still in its infancy, the monolithic microwave integrated circuit (MMIC) differential oscillators [109, 110], high-efficiency (class-E) [111], and C-band high-dynamic range low-noise amplifiers [110], GaN-based MSM planar integrated varactors 6372 (c) Figure 25. (a) Scaled 1/f noise spectral density (SID /ID2 ) before and after high-reverse current stress measured at (VGS − Vt ) = 2.4 V and VD = 0.1 V [127]. (b) Scaled 1/f noise spectral density (SID /ID2 ) versus (VGS − Vt ) at 30 Hz and 1 kHz before and after hot electron pressure [127]. (c) Scaled 1/f noise spectral density (SID /ID2 ) versus (VGS − Vt ) at 30 Hz and 1 kHz before and after high-reverse current stress. The figure shows an increase of SID /ID2 ) in strong accumulation [127]. [112] are reported. An oscillator-based study of degradation of dc characteristics and change of flicker noise due to hot electron and high-reverse current stress in Si3 N4 passivated GaN MODFETs (figure 26) has been done in [127]. It was observed that during hot electron stress, electron trapping in the Wide gap semiconductor microwave devices (a) (b) (c) Figure 26. Structure of the GaN MODFETs [127]. (a) Transfer characteristics before and after hot electron stress, showing degradation of drain current and positive shift of VT plus skewing of transfer characteristics [127]. (b) Logarithmic plots of the transfer characteristics before and after hot electron pressure. The sub-threshold slope decreases after stress and the kink in the characteristic shows the creation of a dominant interface trap. The gate–drain characteristics, in the inset, do not show much change after stress [127]. (c) Transconductance characteristics before and after hot electron pressure. The figure shows degradation of transconductance and positive shift of Vt [127]. barrier layer and interface state creation occur. These cause a positive shift of Vt , reduce ID , skew the transfer characteristics and degrade gm (figure 26). Flicker noise measurements show that after hot electron stress, the scaled drain current noise spectrum (SID /ID2 ) decreases in depletion, but increases only slightly in strong accumulation (figure 25(a)). During high-reverse current stress, electron trapping dominates for AlGaN/GaN HEMT with a 0.5 mm gate width and a 1.1 µm field-plate extension and delivers 1.9 W output power with dc-to-RF efficiency of 21.5%, when biased at Vds = 40 V and Vgs = −4.5 V (figure 28). The resonator was formed by lumped LC components. The small-signal properties were characterized in the frequency range from 50 MHz to 25 GHz. The GaN HEMT MMTC fabrication started with source and drain ohmic contacts. After that, HEMT devices were completed with mesa isolation and gate metallization, as well as passivated by a PECVD SiNx layer, which is also used as dielectric material in MIM capacitors. The chip size is 0.72 mm × 0.7 mm. With a 1.1 µm long field plate, the oscillator delivers 32.8 dBm output power with Figure 27. Circuit schematic of the oscillator [119]. 20% dc–RF efficiency (figures 27 and 28) corresponding to an output power density of 3.8 W mm−1 . The MMIC differential oscillator oscillates at a frequency of 4.16 GHz and provides 22.9 dBm of power from one side at a biasing of Vgs = −1 V and Vds = 20 V (figure 29). Each HEMT has a 0.7 µm × 200 µm gate, the oscillator efficiency 6373 V V Buniatyan and V M Aroutiounian (a) Figure 28. Output power and dc-to-RF efficiency of oscillator with a 1.1 µm long field plate [109]. (b) Figure 30. (a) Schematic of the class-E power amplifier [111]. (b) Simulated voltage and current wave form of the power amplifier [111]. Figure 29. Measured single-side oscillator power of 22.9 dBm from a spectrum analyzer. Vgs = −1 V, Vds = 20 V [110]. is between 4% and 9.4% depending on bias. The output power was found to be comparable to other GaN oscillators (figure 29), while being at least 10 times larger than that of differential oscillators in other technologies of similar topology and size biasing (table 15 [110]). The first AlGaN/GaN HEMT class-E power MMIC amplifiers are reported [111]. The circuit schematic of the amplifiers is presented in figure 30. The circuit achieved a maximum PAE of 57% and an output power of 37.2 dBm at 1.9 GHz and 30 V drain bias. The power density reaches 5.25 W mm−1 at 30 V drain bias and 7.4 W mm−1 at 40 V drain bias. The integrated C-band lownoise HEMT amplifier offers 1.6 dB noise and 10.9 dB gain at 6 GHz (figure 31). With respect to other technologies, it has noise specifications comparable to higher-dynamic range and survivability. With the HEMT based MMIC, there has also been an interest recently in extending the applications to other RF/microwave circuits. Varactors are also essential for tunable circuits. In addition to the capacitance tuning range, quality factor (Q factor) is an important figure of merit for the varactors operating at RF and microwave frequencies. Parameters of the MSM planar inter-digitated varactors, fabricated by the standard HEMT process, have been studied in [112] (figure 32). Varactors have a wide tuning range and exhibit a high quality-factor at 6374 (a) (b) Figure 31. (a) Circuit schematic of the LNA, (b) single tone output power and gain of LNA at 6 GHz [110]. both the maximum and minimum capacitance values. The elimination of ohmic contact resistance in the MSM varactor configuration pushed up the peak Q-factor to 92 at 0.5 GHz and 41 at 1.1 GHz. In conclusion, we note that the dramatic improvement of the crystalline quality of nitride technology and the realization of p-type conduction gave a thrust to research in nitride semiconductors. The breakthrough has led to manufacturing Wide gap semiconductor microwave devices Table 15. Comparison of the oscillator with MMIC, FET-based, oscillators in GaN and differential oscillators in other technologies [110]. Car. freq. (GHz) Gate width (mm) Fund. pow. (dBm) 2nd har. (dBm) 3rd har. (dBm) GaN HEMT Colpitt GaN HEMT Hartley GaN HEMT VCO 5.0 0.2 20.5 −11.5 — 9.56 1.5 32.3 — 9 ± 0.5 1.5 31.8 10.8 [110] 4.16 0.4 25.9 GaAs MESFET Si FET 6.44 4.7 — 0.286 4.67 −9 Best eff. % Phase noise 100 kHz (dBc Hz−1 ) 1 MHz (dBc Hz−1 ) 14.1 −105 −123 — 16 −87 −115 4.8 21 −77 — −19.2 <−55 9.4 −86.3 −115.7 — — −93 −90 −112 −110 and investigations of different types of microwave devices, high-performance blue, green and white LEDs, FETs and UVPhDs. All these devices are robust and will potentially enable a tremendous saving of energy. The availability of ohmic contacts, interconnects and packaging that could function over an acceptable time period at T = 300–600 ◦ C could permit the initial realization of some useful WBG devices and circuits. Given sufficiently reliable contacts, packaging and interconnect, SiC and/or GaN versions of HEMTs, MOSFETs, as well as discrete devices can be used to achieve T = 300–600 ◦ C WBG in discrete devices and MMICs. Robust circuit designs that accommodate large changes in device operating parameters will also be necessary for designing wide temperature range circuit functionality. Unfortunately, obtaining such behaviour from gate insulators on WBG semiconductors has proved to be extremely problematic. Significant technology challenges remain to be overcome for both SiC and GaN, but SiC appears closer to beneficial high ambient temperature functionality than GaN. Since the theoretical high-power switching properties of the WBGs are unmatched at all temperatures, WBG semiconductor devices will play a significant role in realizing beneficial high ambient temperature power conditioning circuits. For low-power circuits, wide bandgap electronics will likely be relegated to the temperature range beyond the reach of SOI electronics, which appears to be above 300 ◦ C. With further improvements of power and noise performance, as next-generation microwave power devices, GaN HEMTs will be an ideal candidate in microwave source applications. The harmonic performance of the GaN oscillator is better than similar differential oscillator design in other technologies. The power amplifier on GaN exhibits state-of-the-art efficiency performance with significant improvement in output power and power density. The GaN/AlGaN heterostructures grown on SiC substrates have a lower level of 1/f noise and higher electron mobility compared with samples grown on sapphire under identical conditions. The AlGaN/GaN HEMTs are promising candidates for high-voltage switching device applications. GaNCVD SiN passivated and insulated-gate AlGaN/GaN HFETs have a potential for high-power operation in the millimeter wave frequency range, especially in the V- and W-bands. Today, the performance of these devices is still being improved, ∼ ∼2 Figure 32. Cross section of the MSM varactor and its physically equivalent circuit [108]. following the steady progress in the areas of crystal growth, processing and fundamental physics. 4. Diamond As noted in the introduction, because of their superior properties, diamond and diamond-like semiconductors promise to be attractive materials for high-power, hightemperature, high-frequency [7, 151–170], radiation induced conductivity modulation [149], optoelectronics [150] as well as surface acoustic waves [151], surface conducting [152], Schottky junction [159], heterojunction [161], metal-intrinsic semiconductor–semiconductor [160], field emission [163, 164], power switching devices, MISFETs [149, 153–158], MESFETs, JFETs, RF MEMS, BJTs [165–168] device applications, for high-power and long-pulse millimeter wave transmission [167] applications, for single-electron transistor [169–172], wear-resistance [173], antireflection coating for solar cells [174] and other applications. If the problems of dopant control and defect density can be overcome, diamond may be widely used for specialized microelectronic applications. In addition, the high thermal conductivity allows the in situ integration of an efficient heat spreader and the low dielectric constant will reduce capacitive loading at high frequencies. The exceptional material properties of diamond are shown in table 16 [7]. Rapid progress has been achieved in the last decade in diamond film chemical vapor deposition (CVD). Since CVD processes and the apparatus for diamond films were established over two decades ago [7, 162–180], many studies have been carried out to understand and improve their material 6375 V V Buniatyan and V M Aroutiounian Table 16. Some exceptional properties of diamond [7]. Property Value Comparison Mechanical hardness (GPa) Thermal conductivity (W cm−1 K−1 ) Fracture toughness (MPa V−1 m−1 ) Young’s modulus (GPa) Coefficient of thermal expansion (ppm K−1 ) Refractive index Transmissivity Coefficient of friction Bandgap (eV) Electrical resistivity ( cm) Density (gm cm−3 ) 80–100 SiC : 40 5–20 Ag : 4.3, Cu : 4.0, BeO : 2.2 5.5 SiO2 : 1, SiC : 4 1050 SiC : 440, Graphite : 9 1.2 SiO2 : 0.5 2.41 at 590 nm 225 nm—far IR 0.05–0.1 (in air) 5.4 1012 –1016 Glass : 1.4–1.8 — Teflon : 1.05 3.51 Si : 2.32, Cu : 89 Method HFCVD MPCVD Dc arc-jet CVD Deposition rate (µm h−1 ) Substrate temperature (◦ C) Deposition area (cm2 ) Advantages 0.1–10 0.1–10 300–1000 300–1200 800–1100 5–900 5–100 Disadvantages Si : 1.1, GaAs : 1.43 AlN : 1014 properties. As CVD is an atomistic process, when the coat grows molecule by molecule, the consequence is that the process is slow and coatings are thin, but the result is a dense high-quality deposition with good adhesion to the substrate due to atomic bond formation. Conventionally, the substrate is thermally activated to initiate the CVD reaction (>800 ◦ C). This could also be a serious limitation to some less stable substrates. Unfortunately, most of the R&D efforts were limited to the laboratory scale and less emphasis was placed on scale-up technology. Many issues, including scale up, cost reduction, reproducibility, difficulties with the preparation of n-type diamond with a high-electron concentration, formation of ohmic and Schottky contacts, etc are still a barrier for practical applications. Successful attempts to use CVD diamond films in diamond electronic devices are presented. Results were obtained on high-temperature high-pressure (HTHP) synthetic single crystal stones. Although it will not allow wafer scale device manufacturing, the stones are already of chip size and their use may still be an attractive approach to demonstrate a discrete high-power diamond device [7]. The mobility of holes and electrons in CVD diamond can be as high as in natural or HPHT synthetic diamond but generally µe and µp are significantly smaller in PCD and HOD due to energy barriers at grain boundaries. Almost all physical properties of diamond are dependent on its crystal structure. Single crystal diamond shows the maximum values for all properties, which then fall off as one moves from polycrystalline to microcrystalline and finally to amorphous forms of carbon. For example, the hole mobility at room temperature for homoepitaxial CVD diamond films is about 1470 cm2 V−1 s−1 . In comparison, it is 229 cm2 V−1 s−1 in highly-oriented diamond film and 70 cm2 V−1 s−1 in polycrystalline film [7]. In small grain CVD diamond the optical and electronic properties are governed by amorphous carbon-like density states located at the grain boundaries. In large grain CVD diamond they approach the properties of a single crystal 1D or 2D diamond. Therefore, the properties are widely spread. The dark conductivity at high temperatures is activated with energies between 1 and 1.7 eV. The activation energy decreases significantly towards lower 6376 Table 17. Comparison of different poly-C deposition methods [7]. 30–150 <2 Simple, Quality, Quality, large area stability high rate Contaminations Rate Contaminations fragile filament small area temperatures. It has been concluded from experiments that nominally undoped CVD diamond is an n-type semiconductor. The typical drift length of charge carriers is in the range of the grain size. EPR experiments [175] allowed the detection of carbon defects [HI] at grain boundaries and nitrogen (PI) in the range of 1017 –1018 cm−3 , which is the origin of the n-type character of CVD diamond. Irrespective of specific applications of CVD diamond films, a large size, production-type CVD apparatus is necessary to upgrade the present CVD technology. Several efforts have already been made to develop a diamond CVD apparatus and processes for high deposition rate and large area coverage [175]. Alternative activation techniques have recently become available which can effectively lower the reaction threshold temperature down to ∼100 ◦ C. These are laser-assisted CVD (LCVD) [7], plasma-activated CVD (PACVD) [176], electron beam inducted CVD (EDCVD), and microwave plasma enhanced CVD (MPECVD) [163, 173, 175–180]. All CVD techniques for the manufacture of diamond films require means of activating gas-phase carbon-containing precursor molecules. This generally involves thermal (e.g. hot filament) or plasma (dc or RF) activation, or use of a combustion flame (oxyacetylene or plasma torches) [7] . While each method differs in details, they all share common features. But the synthesis of diamond at low pressures by CVD techniques is of interest also for significant commercial potential. The low pressure CVD approach is the only realistic alternative to high-temperature high-pressure (HTHP) synthesis methods. Comparison of dc arc-jet discharge, hot filament and microwave plasma enhanced CVD techniques, regimes and parameters are given in table 17 [7]. Another important issue for characterizing CVD diamond films comes with forming ohmic and Schottky contacts. Investigations of contact behaviour of various metals on ntype nitrogen doped ultra nanocrystalline diamond (UNCD) thin films [181] and characterization of diamond ohmic and Schottky contacts is performed in [159, 182]. It was shown that near-ideal ohmic contacts are formed in every case, while Schottky barrier contacts prove more elusive. It is stated that for ‘p-type surface’ conductive diamond films, Au forms ohmic contacts and all-near-ideal Schottky contacts. These contacts play a key role in the overall device performance. Werner [183] reviewed the dependence of the SBH on the surface termination and the impact of annealing of carbideforming metals on the specific contact resistance to diamond. It was concluded that carbide patches dominate the specific Wide gap semiconductor microwave devices contact resistance after annealing. The ohmic or rectifying contact behaviour depends drastically on surface termination. The contact resistance depends strongly on the doping level. In p-type diamond, doping levels much higher than 1020 cm−3 can be easily achieved; therefore the formation of ohmic contacts to diamond is a rather simple operation. It is also well known [159] that hydrogen termination of diamond leads to p-type high-conductivity layers (HCL) near the surface. Contact annealing of carbide-forming metals leads to the formation of carbides at the metal–diamond interface, which reduces the contact resistance. Contact resistances as low as 10−7  cm2 can be achieved after annealing. It was shown that the carbide must not necessarily be a complete layer, but can consist of small diameter carbide patches, which, in turn, have a significant impact on the specific contact resistance, as they have a lower barrier height. It can be expected that the site and density of the carbide patches will depend on the reactivity of carbide-forming metal, annealing temperature and the number of crystal defects on the surface [11]. However, good ohmic contacts to n-type diamond are not so easily understood. The carbide formation does not have the same effect compared with the p-type material. The reported contact resistances are extremely high (>103  cm2 ) and are not useful for practical applications. Polycrystalline metals that include Pt or Pd are not a proper choice. In contrast, amorphous diffusion barriers look much more promising. Another serious hurdle is doping, the most important precondition for the manufacture of electronic devices. Boron is the only deep acceptor partially activated at room temperature. N-type doping by nitrogen or phosphorus is not effective enough to produce high-electron densities and its use in transistor structures is very restricted. The activation energy of boron decreases with an increase in the doping concentration and becomes negligible for net concentrations of NA
1020 cm−3 . Therefore it is possible to fabricate boron doped layers with full carrier activation and a low temperature dependence of their conductivity. 4.1. Diamond based devices As diamond shows theoretically the best performance at high temperature, power and frequency operations of devices (for diamond, at least 40–50 times better than for any other semiconductor), this makes diamond very attractive for RF and microwave electronics applications. The comparison of the properties of diamond with those of other materials shows [7] that: (i) Fmax is theoretically 53 times better for n-type diamond than for Si (42 times in the case of p-type); (ii) Maximal breakdown voltage of a diamond power device is 514 times higher than for Si and 34 times higher than for GaN; (iii) thermal conductivity of diamond is five times larger. Although research is still at the R&D level, some promising results have already been obtained and even some diamond devices (SAW and FE devices) have been commercialized. When combined with diamond high dielectric strength, low dielectric constant and chemical inertness, diamond is seen as an attractive packaging material for microelectronic devices, such as laser diodes and high-power FETs. Diamond is especially attractive for infrared optical utilities because of its broad spectral transmittance at long wavelengths. As a result of its negative electron affinity characteristics, diamond is an attractive candidate material for cold-cathode electron emission applications, including emitter arrays used for flat panel displays. 4.2. Diamond BJT and FETs The first diamond BJT reported in [184] was fabricated by ion implantation and had a low-current application factor of 0.11. In [185] p+ np bipolar junction transistor structures were fabricated by MPCVD growth using nitrogen (base ∼200 nm thick) and boron. Studies showed that the fabrication of abrupt p–n junctions with a high built-in voltage is feasible despite deep activation energy levels of impurities. The analysis has shown high series resistance of the semi-insulating base which will limit BJT operation. The transistor action could only be observed in the nA-current range. The dc current gain of approximately 200 was measured in commonbase configuration, and about 3 in the common-emitter configuration, mainly because of the higher base–collector reverse bias that causes a higher leakage current. Due to the temperature activation of leakage current, the transistor operation was only observed up to 200 ◦ C. Due to the ideal diamond material properties FETs on diamond should theoretically outperform FET structures made of WBG materials like SiC and GaN in high-power RF applications [7, 153–158]. The difficulties in the technology of these structures results in two device concepts investigated in [186]: (a) (i) Boron δ-doped channel JFET, (ii) Boron δ-doped channel MESFET. (b) (i) Surface channel MISFET. (ii) Surface channel MESFET. As the channel depletion in FETs depends on ionized and non-ionized impurities [187], a channel with deep impurities will yield low currents due to incomplete ionization, but at the same time the electric field strength necessary to deplete the channel will be the same as in the case of full activation. The full activation of the boron acceptor at room temperature was obtained in diamond only for the doping concentration NA 1020 cm−3 . This means that the sheet charge and therefore the thickness of the channel will be limited by the breakdown field, thus leading to δ-doped channels. With respect to δ-channel FETs, the δ-channel MESFET with a Schottky-gate has been investigated in [188]. The potential advantages of diamond p–n-junctions in diamond δ-FET structures over Schottky contacts are a lower leakage current and high breakdown as well as the higher built-in voltage of above 3 V compared with the 1.7 V of the Schottky contact. This leads to a shift in the pinch-off voltage towards enhancement-mode operation. The gate metal in the JFET is separated from the channel by an N-doped layer. The N-doped layer is a lossy dielectric represented by a leakage resistor parallel to the dielectric capacitance inserted in between the gate metal and the channel. Depending on the layer properties and dimensions, the behaviour can be either resistive or capacitive. 6377 V V Buniatyan and V M Aroutiounian Figure 33. RF gain plot with extracted cut-off frequency and maximum frequencies of oscillation for LG = 0.2 µm [191]. There are two possibilities to place the gate control junction—on top as conventionally done or at the channel backside [188, 189]. In the surface channel FETs [189], a hole conducting 2DHG-like channel is formed near the surface of hydrogen terminated diamond without extrinsic doping [134, 136]. Today two different structures with two different gate configurations are realized: the surface channel MESFET [148, 190] and the surface channel MISFET [165]. Although the physical/chemical nature of such structures is still under discussion, this type of FET is the only one which demonstrated cut-off frequencies in the GHz range. Smallsignal RF performance of FETs was measured in the frequency range from 1 to 26 GHz [191]. The extracted frequencies for LG = 0.2 µm and WG = 200 µm were fT = 24.6 GHz, fmax(MAG) = 63 GHz and fmax(U) = 80 GHz (figure 33). These values are the highest for diamond FETs at present. The high ratio fmax /fT indicates a high geometrical aspect ratio of the structure. Essentially higher cut-off frequencies are expected in cases where the channel mobility can be improved. First noise measurements have resulted in a minimal noise figure Fmin of ∼1 dB at 3 GHz. Power measurement at 1 GHz indicated a saturated output power density of 0.35 W mm−1 [191]. Maximal drain current of hydrogen-induced surface conductive channel diamond MESFET reaches above 300 mA mm−1 and the maximal drain bias up to 68 V giving a possible RF output power of up to 3 W mm−1 . The RF tests of this device were limited by small signal parameter measurements at low drain current due to unstable performance at high current levels. The maximum cut-off frequencies are fT = 11.5 GHz, fmax(MGH) = 31.7 GHz and fmax(Uf) = 40, 22 GHz. The high fmax /fT ratio of above 3 is related to the high voltage gain of above 100 obtained from the gm /gDS ratio. Such high fmax values enable the millimeter-wave operation, but dc measurements were not accomplished due to instability problems caused by the surface. The maximal transconductance of 100 mS mm−1 , cutoff frequency of 11 GHz and maximal frequency of oscillations of 18 GHz were obtained in the 0.7 µm gate-length diamond MESFETs fabricated on the H-terminated diamond surface [182, 191]. A small-signal equivalent circuit for this device was developed, by matching the measured S-parameters with the calculated ones. The analysis indicates that the 6378 reduction of parasitic resistance between the source and gate is necessary for realizing higher output power. Nippon Telegraph and Telephone Corporation (NTT) have developed a diamond semiconductor device that operates at 81 GHz [192]. The advance promises to make applications in the millimeter-wave band from 30 to 300 GHz possible for the first time. A metal/intrinsic semiconductor–semiconductor FET (MiFET) was fabricated [160] on a Si3 N4 substrate using polycrystalline diamond films grown by MPCVD. The device demonstrated modulation of the drain current for VG  6 V. A maximum value of gm was determined to be 5 µs mm−1 at VD = −10 V and VG = 2 V. However, no saturation of the drain current was observed, which is most likely due to a thin depletion layer and a large gate leakage current. Modulation of the drain current was observed at 77 up to 398 K. The contact resistivity of 10  cm2 between the Ti/Au electrode and the B-doped channel layer is too large compared with the contact resistivity of 10−6 –10−7  cm2 for Si and GaAs. Therefore, the fabrication processes need to be further developed to improve the FET characteristics. 4.3. Diamond based RF MEMS, SAW and field emission devices Doping limits and bandgap engineering in WBG semiconductors, their epitaxial growth, solving of problems of manufacture of ohmic contacts to WBGs, plasma and dry etching, ion implantation are at the centre of researchers’ attention [5]. As mentioned above, diamond and some diamond-like semiconductors are also very interesting. The unique combination of material properties of diamond, such as piezoresistivity, mechanical hardness, low coefficient of friction and high thermal conductivity makes it ideal for applications such as physical sensors and MEMS. The wide gap of diamond indicates that the piezoresistive effect is ‘hard’, that means, it is preserved at higher temperatures and in radiation environments that severely limit present materials [7, 187, 193, 194]. The design of RF circuits is based on a common reference of 50  for interconnects and wiring. In power devices it may deviate essentially from this value in their output and input impedance. Thus, transistor structures are usually surrounded by passive matching networks acting as impedance transformers. All passive components of circuits need to be placed on insulating substrates to avoid losses. In monolithic integration the semiconductor needs semiinsulators which opens problems with the use of the Si and MEMS technologies for passive components. Air gaps can be used for isolation. Diamond can be an ideal insulator with low losses even at millimeter-wave frequencies [195]. Due to its extremely high Young’s modulus, diamond submicron structures may be possible to use as resonators in the GHz range. The electrical switch is an essential part of many electronic systems [7, 149]. It can be used for the switching of microwave power in radar systems and a low loss switch in high resolution measurement system to fuses and circuit breakers. This would open up prospects for ultra-highpower high-temperature electronic systems. Investigations of micro-switches made of polycrystalline diamond films carried out in [196] showed that diamond improves the Wide gap semiconductor microwave devices Figure 34. Left: absolute value of the transmission in the on-state of a diamond microwave relay; Right: absolute value of the transmission in the off-state; length of the relay is equal to 2500 µm [197]. transient behaviour of the devices in comparison with other materials. The results indicate that for operation in air a diamond micro-switch exhibits an approximately eight times higher maximum frequency of operation in comparison with a silicon micro-switch of identical geometry. The potential of the diamond RF MEMS (namely microwave switches in a coplanar arrangement) is reviewed in [197, 198]. Two driving concepts have been evaluated. First is the conventional electrostatic drive operating by a capacitive force, without dc losses. Disadvantages of this concept are high driving voltages and low-contact force. Alternatively, the bimetal effect has been employed in an electro-thermal driving concept. In this concept, the cantilever is bent by thermally induced stress. The actuation voltage can be as low as 1.3 V. However, thermal power is applied to close the switch and is also needed to keep it closed. To investigate the highfrequency behaviour of the relay at microwave frequencies, S-parameter measurements from 45 MHz up to 10 GHz have been performed [197]. The absolute values of the transmission characteristics (S21 ) in the on-state (left) and in the off-state (right) of the relay are shown in figure 34. The transmission in the on-state characteristics shows an attenuation of ∼3.4 dB at 10 GHz. At low frequencies (45 MHz) the relay shows an attenuation of ∼1.2 dB. The attenuation in the off-state is about −23 dB at 10 GHz. The characteristics of the conductivity modulation of diamond switches induced by continuous electron beam excitation are investigated in [149]. The current gain depends strongly on the energy and current of the electron beam. At low electron beam currents, continuous gain up to 180 has been measured for 30 keV electrons. The results also show that the electroconductivity gain obtained in these conditions is much higher than the UV photoconductive gain and increases with electron energy. It seems that the most promising area for future development will be in the field of microelectromechanical structures, as their friction, striction and wear properties make them prime candidates for use in moving mechanical assemblies. Other attractive applications of diamond are surface acoustic wave (SAW) devices [151, 199, 200, 201], because diamond has the highest sound velocity among all materials, and field emission [7, 199, 202, 203]; due to such field emitter arrays (FEA), direct density modulation of the electron beam at RF frequencies becomes possible. For example, the 5 GHz diamond SAW filter has been fabricated with an electrode of 0.5 µm width, whereas electrodes of less than 0.2 µm are necessary for conventional SAW materials [200]. In addition to the highest Young’s modulus, diamond also has the highest thermal conductivity. This will also provide an advantage for high-power handling of SAW devices. Since diamond does not possess piezoelectricity, it must be combined with a piezoelectric layer like the ZnO film not only for the generation of the SAW but also for the transparent ZnO/diamond heterojunctions (n-type ZnO with p-type diamond) [161,204]. Some companies are already exploiting diamond-based SAW filters in mobile phone equipment, and it is likely a diamond SAW filter will be an essential component of all high-frequency communications equipment, including telephone networks, cable television and Internet. Due to low electron transit time and high transconductance in a triode structure with grid gate, making it possible to lift an upper modulation frequency to more than 2 GHz, the diamond NT based electron emitters are suitable for use in a novel type of microwave amplifier [205] capable of producing of order 10 W at 30 GHz. These devices are vacuum tubes and should not be confused with the solid-state power devices. The application area will be the long-range telecom systems that are based upon microwave links. Kang and coworkers have reported the development of vertical and lateral diamond field emission devices [7, 203]. The vertically self-aligned gated diamond vacuum triodes fabricated on a silicon-on-insulator (SOI) [203] show high cathode current potential. The field emission of the triode array exhibits transistor characteristics with high dc voltage gain (800) and has good transconductance. The lateral diode exhibits a low turn-on voltage of 5 V (field ∼3 V µm) and a high emission current of 6 µA. The lateral diamond field emitter has potential applications in vacuum microelectronics, sensors, MEMS, field emission displays, electron microscopy etc. Below, a summary of the current status of suitability of diamond for a technology choice in RF applications and some problems are discussed [7]. • Bipolar junction transistors (BJTs) – Deep doping problem – High resistivity of the nitrogen doped base (10 G cm at 20 ◦ C) 6379 V V Buniatyan and V M Aroutiounian • • • • – nA current range – Operation temperature below 200 ◦ C (leakage currents dominate) – Base transport is around 100 GHz due to high carrier mobility – With nitrogen doped base the only area will be high dc current gain applications – 50–70 GHz cut-off frequencies – For accurate modeling, high series resistance of the base, leakage of reverse bias p-NT junctions and surface leakage must be examined – Common base configuration; β is up to 200, where it is 1.1 in common emitter configuration Field effect transistors (FETs) – Maximum drain bias of 200 V is demonstrated – Power density of nearly 30 W mm−1 is extracted (measured value is 0.35 W mm−1 at 1 GHz, LG = 0.4 µm with 400  load) – fT = 21 GHz and fmax = 63 GHz is measured (for 0.2 µm gate length) – Passivation issues have not been addressed yet – Most promising active devices Passive components and MEMS – RF filters of high Q with 640 MHz resonance frequency obtained – Diamond contacts avoid sticking – High-power and high-speed switching with diamond contacts – Coplanar waveguides have been realized – Elementary switching of MW signals in a coplanar waveguide arrangement has been demonstrated – Diamond RF MEMS and all diamond structures based on membranes and free standing beams are attractive Surface acoustic wave (SAW) devices – First commercial diamond electronics product – Promising for high-frequency equipment – Likely to replace conventional SAW filters Field emission (FE) devices – Oscillation from dc to 100 THz obtained – Promising for long-range telecom systems 4.4. Carbon NT based structures As stated, carbon nanotubes (CNTs) have excellent electrical properties. Moreover, NTs have extraordinary mechanical strength and high thermal conductivity. They can act as onedimensional ballistic conductors at room temperature as well as field effect transistors with performance comparable to that of silicon FETs [152, 165, 168]. Recently, new functionalities of NT devices have been demonstrated experimentally, e.g. gate controlled IR electroluminescence from FETs [206]. As the continuous downscaling of CMOS technology will meet serious challenges within the coming decade, it is time, in addition to the performance of individual NT devices, a number of critical issues regarding the integration of NTs with existing semiconductor technology were discussed. The more realistic approaches will be hybrid schemes for the integration of NTs with contemporary technology for enhanced performance of specific components [198]. Such integration 6380 is presently unfeasible due to a number of unsolved critical issues. Such problems concern material capabilities, electrical contacts, functionalities, circuit architecture, etc. The fabrication of semiconductor devices involves numerous processing steps, including metallization, oxide deposition, lithography and etching. Investigations carried out in [161, 162, 175–180, 198, 207–209] showed that NTs can be incorporated directly into a semiconductor matrix grown by MBE, which is performed with III–V semiconductors. CNTs endure temperatures up to 700 ◦ C and are therefore compatible not only with III–V growth but also with both MBE and CVD epitaxial growth of Si and Ge. It is also concluded that metal electrodes can be evaporated onto the NTs and dielectric layers can be deposited. The closed sp2 bonded carbon structure of the NT surface makes them inert to most standard chemical reactants and solvents used in processing. CNTs are not affected by exposure to UV or electron beam lithography and, moreover, etchants such as hydrofluoric acid used for SiO2 removal do not harm the electronic properties of CNTs. Conversely, NTs can be oxidized and etched almost selectively by oxygen plasma treatment [173, 176–178], which is thus useful for removing CNTs from larger areas in combination with masking by resistor solid material. For the formation of contacts to CNTs on the Si substrate, a range of different electron materials has been tested and tunnel barriers and transparent contacts have been fabricated [208]. The NTs can even be contacted with superconducting or ferromagnetic electrodes. Since NTs are one-dimensional conductors, their contact resistance when coupling electronically with two- or three-dimensional contacts will however exceed h/4q 2 = 6 k (for single wall CNT), which means a fundamental mismatch with the 50  impedance standard for current semiconductor circuits and this may potentially become a problem for applications. The fabrication of NT devices on the top of piezoelectric substrates can also be realized. For the integration of NTs with silicon MOS circuits, the semiconductor substrate circuit was achieved through MO metal pads on the top of the integrated circuit [202], which was completed before incorporating NTs directly into a host crystal by epitaxial overgrowth during the semiconductor fabrication procedure [198]. Successful incorporation of NTs inside semiconductor materials opens possibilities for direct coupling to bandgap-engineered structures such as quantum wells or optical cavities for optoelectronics. The NTs with III–V semiconductors have been integrated by MBE [210]. The most important NT component in the context of electronic applications is FET based on semiconducting NTs [170–173, 208, 211–216]. Key characteristics of the three main types of CNFET are compared to the Si MOSFET having the smallest gate length and oxide thickness in table 18 [173]. The electrolyte gate CNFET is the best when it comes to transconductance, with the metal top gate as the second one. Both CNFETs are beating the Si MOSFET already. Moreover, they tie with Si MOSFET in on/off ratio. Given these two factors, despite the thicker gate oxide, means that the CNFETs have the potential to easily outperform Si MOSFET technology in the near future [154]. Today CNFETs are similar to the MOSFETs. The main difference between CNFETs and MOSFETs is that CNFETs Wide gap semiconductor microwave devices use their CNT as the channel; whereas a MOSFET’s channel is made of silicon. Both technologies use complementary p-type and n-type devices. This helps to provide reduced power consumption, higher gain, stability and easy implementation of logic circuits. The first CNFETs were designed in the simplest ways (figure 35). The NT is draped over two gold electrodes, representing the source and drain. The gate is located either on the side of or underneath the transistor. In addition, the gate is separated by an insulator, such as silicon dioxide, to isolate the gate from the NT. By having it isolated, the electric field from the gate is able to switch the transistor on and off. If it were not shielded, a short would occur and the transistor would not be able to operate. There are problems with these designs. The first is that the NT is left open to the air, which, as we will see, causes the transistor to operate as a p-type or p-FET only. In addition, the gate placement causes problems. The gate oxide is required to be thick (at least ∼100 nm) which causes the need for high voltages to penetrate the oxide and switch the transistor. The second generation transistors boast a big improvement over the first design [172, 211–216]. The main difference between the first and the second generations is the placement of the gate electrode. The gate is now placed on top of the NT. This design change seals the NT, so the air will not force the transistor to act as a p-type transistor. The other benefit is that the gate oxide is much thinner (about 15 nm), allowing one to use much smaller voltages. In fact, the only difference is in the channel. An additional advantage is that the second-generation CNFET can be well suited for high-frequency operations. This is not possible with the first generation because of the large overlap capacitance from all three electrodes. Table 18. Comparison of the different CNFET geometries along with the conventional MOSFET [173]. P-type FETs Back-gated Metal top Electrolyte Si CNFET gate top gate MOSFET Gate length 1030 (nm) Gate oxide 150 thickness (nm) Transconductance 244 (mS mm−1 ) Subthreshold slope 730 (mV dB−1 ) On/off ratio 105 260 — 100 15 — 0.8 2100 130 106 10 000 460 — 80 106 106 Another way to create the second generation design is to use an electrolyte to make the gate (figure 36). The higher capacitance here improves performance. The curves are very similar to the MOSFET curves and to the metal top-gate design. One key difference is that the saturation point is reached earlier than in both the metal top-gate design and the MOSFET. This means that this design might potentially consume less power than both the other transistor types. However, the fact that it uses a liquid electrolyte is a drawback. The electrolyte must be sealed to prevent leakage and to keep the air from diffusing through and converting n-type into p-type, which adds extra steps in manufacturing. The C–V and transfer characteristics for the p- and n-type CNFETs are discussed in [172]. Based on the transfer characteristics, we can see that there is a little bit of conduction when the gate is positive biased; however, this is too small to accidentally turn on another transistor. It will only add a minute amount of power wastage. When the gate voltage is negative, the transistor turns on and delivers a five-fold increase in drain current. This is good because it means that the transistor will easily be able to drive another one. The researchers’ group at IBM compared some properties of the CNFET with both a high-performance silicon MOSFET and a new MOSFET design that utilizes silicon-on-insulator (SOI) technology; the results are displayed in table 19. One important difference is in Ioff . The CNFET possesses a significantly smaller value of current which is approximately 70% less than the conventional MOSFET. This means that power is being wasted while the transistor is off. In addition, we notice that Ion , or drive current, is larger. In fact, it is three to four times larger. Since the NT has ballistic conductance, it actually has a smaller resistance and leads to two to four times increase in transconductance. When CNFET design is optimized, the CNFET will surely outperform both the current technology and the newer SOI technology. Note that for active device applications similar to the conventional FETs, the semiconducting-type single wall NT (SWNT) is more suitable than the metallic-type SWNT because the channel conductance in the former device can be controlled/modulated by the gate (voltage). If the CNTs are multi-well (NWNT) in nature, they will depict metallictype electrical behavior and will not be suitable for future nanoelectronic applications [153]. In general, the CNT-FET acts like a p-type conduction device when it is exposed to air [154]. This phenomenon is ascribed to the absorbed oxygen between the interface of the metal/SWNT, causing the pinning of the Fermi level near the valence band. Once Figure 35. First generation carbon NT FET (CNFET) [170]. 6381 V V Buniatyan and V M Aroutiounian 5. Conclusion Figure 36. Schematics of a CNFET that uses an electrolyte solution for the gate [173]. Table 19. Comparison of CNFETs to both MOSFETs and SOI MOSFETs [172]. p-type CNFET MOSFET SOI MOSFET Gate length (nm) Gate oxide thickness (nm) Vt (V) ION (µA µm−1 ) (Vds = Vgs − Vt ≈ −1 V) IOFF (nA µm−1 ) Subthreshold slope (mV dB−1 ) Transconductance (µS µm−1 ) 260 15 15 1.4 −0.5 2100 ∼ − 0.1 265 150 130 2321 <500 ∼100 975 50 1.5 ∼ − 0.2 650 9 70 650 p-type CNT-FET is annealed in vacuum, the absorbed oxygen in the interface will be removed and the CNT-FETS is transformed from pure p-type conduction to ambipolar or n-type conduction, depending on the annealing condition and/or the amount of remaining oxygen [156–158]. It is worth noting that if a device is ambipolar, it conducts either electrons or holes depending on the gate bias. For the conventional structure with a traditional single gate, the unipolar phenomenon cannot be easily suppressed. However, the electrical characteristics of the converted CNT-FETs vary widely and uncontrollably in this process, and across-the-chip variations of FETs often widen and become intolerable after annealing. The Chen group [153] manufactured air-stable n-type and p-type CNTFETs without any additional or complex annealing process (table 20); it was found that some ambipolar-type devices were manufactured. The type-2 device depicts either n- or p-type characteristics rather than the ambipolar-type, depending on the passivation layer deposited on the SWNTs. It is difficult to alter the conduction-type of individual CNT-FETs located on the same chip, in [153] a novel double gate DG CNT-FET structure which can provide a practical and reproductive method to perform both n- and p-type like CNT FET devices as well as unipolar-type CNT FET devices on the same chip is proposed. It is shown that ambipolar-type CNT-FET can be converted to n- and p-type CNT-FET by controlling the biases applied to the top- and bottom-gate electrodes, which is very important for CMOS circuit applications. Therefore, even though the initial CNT FET shows ambipolar behaviour, the electron flow channel is effectively blocked off by the negative top-gate bias. 6382 Silicon devices and most traditional integrated circuits made of Si are not able to operate at temperatures above 150 ◦ C, especially when high operating temperatures are combined with high-frequency, high-power, and high-radiation environments. The maximum operating temperatures of wide bandgap (WBG) semiconductors are shown in table 1. Considerable interest in recent years in WBG semiconductors and devices is especially connected with two important classes of applications, namely blue/green and white light emitting diodes and lasers as well as high-power/highfrequency/high-temperature electronics. For optoelectronics applications, GaN, GaAlN, AlN and other WBG semiconductors are leading materials because they are direct bandgap materials which give them a huge advantage over the indirect gap SiC, although it has been used for a long time also for the manufacture of LEDs. But it is necessary to mention that lattice mismatches of only 1% for AlN and 3% for GaN exist when these materials are grown on 6H-SiC substrates. Thus, SiC processing is often intimately linked with AlN and GaN electronic and optical device fabrication as well. For example, commercially available GaN on SiC blue LED represented the highest volume of blue LED sold. The small lattice mismatches allow using such combinations of these WBG semiconductors as materials with an immense potential for use in hetero-optoelectronic devices. For the second major class of applications, SiC is the most important material with diamond and GaN as other candidates. Figures of merit for various WBG semiconductors are shown in table 3 and figure 11. Diamond and diamondlike carbon actually have very good parameters, but problems with producing large single crystals and doping delay their applications. GaN and GaAlN have the advantage of the availability of heterostructures and excellent transport properties, but they have poorer thermal conductivity in comparison with silicon carbide. Doping limits and bandgap engineering in WBG semiconductors, their epitaxial growth, solving of problems of manufacture of ohmic contacts to WBGs, plasma and dry etching and ion implantation are in the focus of attention of researchers. Remarkable achievements in technology are fixed. Much attention has been given to SiC, currently the most mature of the wide bandgap (2.0 eV < Eg < 7.0 eV) semiconductors, as a material well suited for high-temperature, high-frequency and efficient high-power operation. Hightemperature circuit operation from 350 to 500 ◦ C is desired for use in aerospace applications, nuclear power instrumentation, satellites, space exploration, etc. SiC presently has several advantages including commercial availability of substrates since ∼1991 and its large size produced today, known device processing techniques, and the ability to grow a thermal oxide for use as masks in processing, device passivation layers and gate dielectrics. In addition, SiC’s high thermal conductivity (about 3.3 times that of Si at 300 K for 6H–SiC) high electric field breakdown strength (about 10 times that of Si for 6H–SiC), and wide bandgap (about 3 times that of Si for 4H–SiC and 6H–SiC) make it a material ideally suited for high-temperature, high-power, high-frequency and Wide gap semiconductor microwave devices Table 20. The electrical characteristics of Type I and Type II devices [153]. κ Type IIb κ Type IIb Type Ib κ The gate dielectric layer of bottom gate/deposition temperature/ film thickness The gate dielectric layer of top gate/deposition temperature/film thickness Majority type of the devices on a wafer PE-oxide/400 ◦ C/200 nm PE-oxide/400 ◦ C/200 nm PE-nitride/390 ◦ C/200 nm PE-nitride/390 ◦ C/200 nm Thermal-oxide/400 ◦ C/100 nm Thermal-oxide/400 ◦ C/100 nm PE-oxide/400 ◦ C/200 nm PE-nitride/390◦ C/200 nm PE-oxide/400 ◦ C/200 nm PE-nitride/390 ◦ C/200 nm PE-oxide/400 ◦ C/200 nm PE-nitride/390 ◦ C/200 nm p-typea n-typea p-typea n-typea p-typea n-typea a Some of these devices contain ambipolar-type characteristics. The major differences between Type-I and Type-II lie in the separate bottom-gate design in Type-II structures and the different dielectric layers between the two types. Note: κ—Data not shown. b high-radiation environments. Besides, excellent microwave parameters are demonstrated in this review for SiC BARITTs, TUNNETTs and BARIQWTTs (quantum dot SiC structures). The use of silicon carbide for the manufacture of TUNNETTs can promise better microwave parameters over the range of frequencies 100–500 GHz. A wide array of devices in 3C–, 6H– and 4H–SiC have been demonstrated, including: the bipolar junction transistor, insulated-gate bipolar transistor, MESFET, JFET, MOSFET (which has been operated up to 923 K), static induction transistor (SIT), as well as power MOSFETs, thyristors, gate turn-off thyristors (GTOs), p–n-junction diodes and Schottky diodes. Additionally, heterojunction devices such as the heterojunction FET (HFET) and heterojunction bipolar transistors (HBTs) have been proposed and fabricated. Examples include 3C–SiC/Si HBTs, 6H–SiC/3C–SiC HBT and GaN/6H–SiC HBT. The SiC p-i-n rectifiers and other power electronic devices with good DC performance have been commercially available for many years. Many SiC devices are being developed for microwave power amplifier and oscillator applications. No doubt, SiC is today the most important material, after GaAs and Si, in microwave semiconductor electronics. Characteristic of nitride based diodes and GaN/SiC heterojunction structures, Schottky barrier heights measured on 4H–SiC, 6H–SiC, GaN and AlGaN are investigated. Recently the AlGaN/GaN FET, GaN HEMT and AlGaN/AlN/GaN HEMT on a SiC substrate working in high-power and highfrequency (up to 155 GHz) electronic circuits are reported. The superior thermal properties of GaN have resulted in the realization of heterojunction FETs operating up to 750 ◦ C. The main attention in recent years has been drawn to SiC MOSFETs and GaN HEMTs, but the development of high-performance SiC MOSFETs has been hampered due to poor MOS channel mobility and reliability. Anyway, the maximum oscillation frequency of GaN HFETs was between 163–192 GHz. The Emode Si3 N4 /AlGaN/GaN MISHFET with a 1 µm gate length and a 15 nm Si3 N4 layer inserted under the metal gate to provide additional isolation between the gate Schottky contact and AlGaN surface, which can lead to reduced gate leakage current and higher-gate-turn-on voltage, had a maximum current density of 420 mA mm−1 . The progress in design and manufacture of diamond BJTs and FETs, diamond based RF MEMS, SAW and field emission devices is obvious. Diamond can be an ideal insulator with low losses at millimeter wave frequencies. Due to its extremely high Young’s modulus, diamond submicron structures may be possible to use as resonators and relays up to 10 GHz range. The diamond nanotube based electron emitters are suitable for the use in a novel type of microwave amplifier capable of producing of order 10 W at 30 GHz. Carbon nanotube FETs are very promising devices which are beginning to be used in the microwave range. Acknowledgments The work of the authors in the field of microwave semiconductor devices was supported for many years within the framework of regular grants in Armenia, ISTC A322 grant, INTAS-94-3912, INTAS-96-268 grants and ANSEF PS32-2003 grant. Now these works are carried out under support from the Armenian Target Program ‘Semiconductor nanoelectronics’ and the ISTC A1232 grant. A part of the investigations was carried out in SIMA of the Saclay Research Center. References [1] Casady J B and Johnson R W 1996 A review Solid-State Electron. 39 1409–22 [2] Baliga B J 1996 IEEE Trans. Electron Devices 43 1717–31 [3] Weitzel C E et al 1996 IEEE Trans. Electron Devices 43 1732–9 [4] Neudeck P G, Okojie R S and Chen L-Y 2002 Proc. IEEE 90 1065–76 [5] Pearton S J et al 1999 J. Appl. Phys. 85 1–78 [6] Soukiassian P G and Enriquez H B 2004 J. Phys.: Condens. Matter 16 S1611 [7] Gurbuz Y et al 2005 Solid-State Electron. 49 1055–70 [8] Chow T P et al 2000 Solid-State Electron. 44 277–301 [9] Akasaki I 2002 J. Cryst. Growth 237–239 905–11 [10] Chernyak L, Osinsky A and Schulte A 2001 Solid-State Electron. 45 1687–702 [11] Ostling M et al 2002 Proc. 23rd Int. Conf. Microelectron. 1 32–9 [12] Sze S M 1985 Semiconductor Device Physics and Technology (New York: Wiley) [13] Padovani F A and Stratton R 1966 Solid-State Electron. 9 695 [14] Hatakeyama T and Shinahe T 2002 Mater. Sci. Forum 389–393 1169–72 [15] Mnatsakanov T T, Levinshtein M E and Pomortseva L I 2002 Semicond. Sci. Technol. 17 974–7 6383 V V Buniatyan and V M Aroutiounian [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] 6384 Mnatsakanov T T et al 2003 Solid-State Electron. 47 111–15 Weitzel C E et al 1996 Solid-State Electron. 43 1732–9 Trew R J 1997 Phys. Status Solidi 162 409–19 Domeij M et al 1999 Int. Conf. on Silicon Carbide Related Materials 1999, Mater. Sci. Forum 338–342 1327–30 Moore K, Weitzel C, Nordquist K, Pond L, Palmour J, Allen S and Carter C 1995 Proc. IEEE/Cornell Conf. Advanced Concepts in Speed Semiconductor Devices and Circuits (Ithaca, NY, August 1995) pp 40–6 Siergej R R et al 1995 IEDM Digest (Washington, DC, December 1995) pp 353–6 Clarke R C et al 1996 Device Research Conf. Digest (Santa Barbara, CA, June 1996) pp 62–3 Sung Y M, Casady J B, Dufrene J B and Agarwal A K 2002 A review of SiC static induction transistor development for high-frequency power amplifiers Solid-State Electron. 46 605–13 Nakamura D et al 2004 Nature 430 1009–14 Bencuya I et al 1975 IEEE Trans. Electron Devices ED-32 1321–5 Weitzel C E et al 1994 IEEE Electron Device Lett. 15 406–8 Zhang A P et al 2003 Solid-State Electron. 47 821–6 Pan J N, Cooper J A Jr and Melloch M R 1997 IEEE. Electron. Devices Lett. 18 349–51 Yu L C and Sheng K 2006 Breaking the theoretical limit of SiC unipolar power device—a simulation study Solid-State Electron. 50 1062–72 Allen S T, Pribble W L, Sadler R A, Alcorn T S, Ring Z and Palmour J W 1999 IEEE MTT-S Digest vol 23, p 321 Moore K E et al 1997 IEEE Electron Devices Lett. 18 69 Noblanc O, Arnodo C, Dua C, Chartier E and Brylinski C 2000 Mater. Sci. Forum 1247 338–42 Aroutiounian V M, Avetisyan G A, Buniatyan V V, Soukiassian P G and Buniatyan V 2006 Appl. Surf. Sci. 252 5445–8 Mehdl I, Haddad G I and Mains R K J 1988 Appl. Phys. 64 1533–40 Tager A S 1974 Vyssh. Uchebn. Zaved. Izv. VUZ-ov Radioelektron. 17 3–18 (in Russian) Aroutiounian V M, Buniatyan V V and Soukiassian P G 1999 Solid-State Electron. 43 343–6 Aroutiounian V M, Buniatyan V V and Soukiassian P 1999 IEEE Trans. Electron Devices 46 585–58 Aroutiounian V M, Buniatyan V V and Soukiassian P 2001 Proc. SPIE 4499 160–6 Buniatyan V V and Aroutiounian V M 1998 J. Appl. Electromagn. Athens. 1 1–11 Aladinsky V K and Gradinov P G 1972 Russ. J. Radio Eng. Electron. 17 376–80 Motoya K and Tunnett M J 1985 Int. J. Infrared Millim. Waves 6 483–95 Ruff M, Mitlehner H and Helbig R 1994 IEEE Trans. Electron Devices 41 1040–53 Yeh S 1985 IEEE Trans. Electron Devices 28 117–24 Buniatyan V V, Aroutiounian V M, Zekentes K, Camara N and Soukiassian P 2004 Solid-State Electron. 48 1569–77 Aroutiounian V M, Buniatyan V V, Soukiassian P and Buniatyan V 2006 J. Phys. IV France 132 137–9 Gribnikov Z S, Vagidov N Z and Haddad G I 2004 J. Appl. Phys. 96 1489 Misawa T 1967 IEEE Trans. Electron Devices 14 795–80 Vassilevski K V et al 2002 Mater. Sci. Forum 389–393 1353–8 Itoh A and Matsunami H 1997 Phys. Status Solidi a 162 389–408 Yuan L, Cooper J A, Webb K J Jr and Melloch M R 2002 Mater. Sci. Forum 389–393 1359–62 Gruzinskis V et al 2002 Mater. Sci. Forum 338–342 1379–82 Molions F et al 2002 J. Appl. Phys. 91 5116–24 Clerc R, Spinelli A, Ghibaudo G and Pananakakis G 2002 J. Appl. Phys. 91 1400–9 [54] Bjork M T et al 2002 Appl. Phys. Lett. 81 4458–60 [55] Aderstedt E and Lundgren P 2002 Solid-State Electron. 46 1033–7 [56] Khan R U and Pal B B 1983 J. Inst Electron. Telecommun. Eng. 29 68–71 [57] Aroutiounian V M, Buniatyan V V and Soukiassian P 1999 Proc. ISDRS. (USA, 1–3 December) pp 359–63 [58] Aroutiounian V M, Buniatyan V V, Soukiassian P, Zekentes K and Buniatyan V 2006 J. Phys. IV France 132 355–7 [59] Ikoma Y, Uchiyama K, Watanabe F and Motooka T 2002 Mater. Sci. Forum 389–93 751–4 [60] Fissel A et al 1999 Mater. Sci. Eng. B 61–62 139–42 [61] Bauer A et al 1999 Mater. Sci. Eng. B 61–62 179–82 [62] Danielsson E et al 1999 Simulation Mater. Sci. Eng. B 61–62 320–4 [63] Chein F R and Nutt S R 1999 J. Appl. Phys. 77 3138–45 [64] Van Zeghbroeck B et al 2000 Solid-State Electron. 44 265–70 [65] Lebedev A A et al 2002 Mater. Sci. Forum 389–393 683–6 [66] Wu K H et al 1999 Solid-State Electron. 43 1215–17 [67] William J et al 2000 GaN/SiC heterojunction bipolar transistors Solid-State Electron. 44 259–64 [68] Aroutiounian V M, Buniatyan V V and Petrosyan S G 1999 Proc. SPIE. 3795 551–60 [69] Aroutiounian V M, Buniatyan V V and Petrosyan S G 1998 Izv. Akad. Nauk. Arm. Fiz. 33 241–8 (in Russian) [70] Bechstedt F et al 2002 Mater. Sci. Forum 389–393 737–42 [71] Meng C C and Harold R 1995 US Patent 5466965 [72] Mokerov V G et al 1999 Russ. Microelectron. 28 1–12 [73] Kaniewska M et al 1996 Electron Technol. 29 131–8 [74] Levine B F 1993 J. Appl. Phys. 74 R1–79 [75] Levine B F et al 1993 J. Appl. Phys. 72 4429–43 [76] Ershov M, Ryzhii V and Hamaguchi C 1995 Appl. Phys. Lett. 67 3147–9 [77] Lee S-K et al 2002 Solid-State Electron. 46 1433–40 [78] Ivanov P A et al 2000 Semicond. Sci. Technol. 15 908–10 [79] Badila M et al 2002 Diamond Relat. Mater. 11 1258–62 [80] Sochacki M et al 2002 Diamond Relat. Mater. 11 1263–7 [81] Bhatnagar M and Baliga B J 1993 IEEE Trans. Electron Devices 40 645–55 [82] Ivanov P A et al 2006 Solid-State Electron. 50 1368–70 [83] Elasser A et al 2000 Solid-State Electron. 44 317–23 [84] Soloviev S and Cherednichenko D 2004 J. Appl. Phys. 95 4376–81 [85] Camara N et al 2006 IEEE Trans. Electron. Devices 27 108–12 [86] Nigam S et al 2003 Solid-State Electron. 47 61–4 [87] Tone K et al 2001 Semicond. Sci. Technol. 16 594–7 [88] Gardner C T et al 1992 Appl. Phys. Lett. 61 1185–9 [89] Cai Y et al 2005 IEEE Electron Device Lett. 26 435 [90] Cheng Z et al 2005 IEEE Electron Device Lett. 26 521–3 [91] Huang H K et al 2003 Solid-State Electron. 47 1989–94 [92] Sanabria C et al 2006 IEEE Electron Device Lett. 27 19–21 [93] Wang R, Cai Y and Tang W 2006 IEEE Electron Device Lett. 27 633–4 [94] Dora Y et al 2006 IEEE Electron Device Lett. 27 713 [95] Palacios T et al 2006 IEEE Electron Device Lett. 27 13 [96] Nidhi A N et al 2006 IEEE Electron Device Lett. 27 877 [97] Wu Y-F et al 2004 IEEE Electron Device Lett. 25 117–19 [98] Saito W et al 2005 IEEE Electron Device Lett. 52 159–64 [99] Parish G et al 2001 Phys. Status Solidi 188 297–300 [100] Tian Y et al 2003 Solid-State Electron. 47 1863–7 [101] Monroy E et al 2001 J. Cryst. Growth 230 537–43 [102] Cao X A et al 2006 IEEE Electron Device Lett. 27 329 [103] Huh C et al 2003 Electrochem. Solid-State Lett. 6 G79–81 [104] Zehnder U et al 2001 J. Cryst. Growth 230 497–502 [105] Bru-Chevallier C et al 2004 Thin Solid Films 450 75–8 [106] Buchheim C et al 2004 Thin Solid Films 450 155–8 [107] Sandvik P et al 2001 J. Cryst. Growth 231 366–70 [108] Chu C S et al 2005 IEEE Electron Device Lett. 26 432 [109] Liu J-Q et al 2005 Solidaire Electron. 49 1961–4 [110] Xu H et al 2004 IEEE Microwave Wireless Components Lett. 14 262–4 Wide gap semiconductor microwave devices [111] [112] [113] [114] [115] [116] [117] [118] [119] [120] [121] [122] [123] [124] [125] [126] [127] [128] [129] [130] [131] [132] [133] [134] [135] [136] [137] [138] [139] [140] [141] [142] [143] [144] [145] [146] [147] [148] [149] [150] [151] [152] [153] [154] [155] [156] [157] [158] [159] [160] [161] [162] [163] Sanabria C et al 2005 IEEE Microwave Wireless Components Lett. 15 463–5 Xu H et al 2006 IEEE Microwave Wireless Components Lett. 16 22–4 Khan M A et al 1993 Appl. Phys. Lett. 63 1214–15 Wu Y F et al 1997 IEEE Electron Device Lett. 18 438–40 Dieci D et al 2000 IEEE Trans. Electron Devices 47 261–8 Mooljii A A, Bahl S R and del Alamo J A 1994 IEEE Electron Device Lett. 15 313–15 Johnson J W et al 2002 Solid-State Electron. 46 513–23 Higashiwaki M et al 2006 IEEE Electron Device Lett. 27 719 Dora T et al 2006 IEEE Electron Device Lett. 27 529–30 Chen C-H et al 2000 IEEE Electron Device Lett. 21 549 Tipirneni N et al 2006 IEEE Electron Device Lett. 27 716–18 Rumyantsev S et al 2000 J. Appl. Phys. 87 1849–54 Yagi S et al 2006 Solid-State Electron. 50 1057–61 Higashiwaki M et al 2006 IEEE Electron Device Lett. 27 16–18 Wang R et al 2006 IEEE Electron Device Lett. 27 793–5 Neudeck P G, Okojie R S and Liang-Yu C 2002 Proc. IEEE 90 1065–76 Huang W et al 2006 IEEE Electron Device Lett. 27 796–8 Sahoo D K et al 2003 IEEE Trans. Electron Device Lett. 50 1163–70 Vertiatchikh A et al 2006 Solid-State Electron. 50 1425–9 Anwar A F M et al 2006 Solid-State Electron. 50 1041–5 Lu W et al 2001 IEEE Trans. Electron Device Lett. 48 581–5 Wu Y-F et al 2004 IEEE Electron Device Lett. 25 117–19 Daumiller I et al 1998 Device Research Conf. Digest 114–15 Kim D W et al 2001 Thin Solid Films 398–399 93–8 Danielsson E et al 2002 Solid-State Electron. 46 827–35 Qiao D et al 2000 J. Appl. Phys. 87 801–4 Danielsson E et al 2002 J. Appl. Phys. 91 2372–9 Shiojima K et al 2002 Appl. Surf. Sci. 190 318–21 Schwbgler V et al 2001 J. Cryst. Growth 230 512–16 Nakamura S et al 1998 Appl. Phys. Lett. 72 211–16 Akasaki I 2002 J. Cryst. Growth 905–11 Xu H et al 2004 IEEE Microwave Wireless Components Lett. 14 262–4 Zheleva T S et al 1997 Appl. Phys. Lett. 71 2472 Roskowski A M et al 2002 IEEE J. Quantum Electron. 38 1006 Zubia D and Hersee S D 1999 J. Appl. Phys. 85 6492 Chernyak L, Osinsky A and Schulte A 2001 Solid State Electron. 45 1687 Qiao D et al 2000 J. Appl. Phys. 87 801–4 Tans S, Verschueren A and Dekker C 1998 Nature 393 49 Aleksov A et al 2002 Diamond Relat. Mater. 11 382–6 Achard J et al 2004 Diamond Relat. Mater. 13 876–80 Saby C and Muret P 2002 Diamond Relat. Mater. 11 851–5 Seo S-H, Shin W-C and Park J-S 2002 Thin Solid Films 416 190–6 Jackman R B et al 1999 Carbon 37 817–22 Chen B-H et al 2006 Solid-State Electron. 50 1341–8 Wind S T et al 2002 Appl. Phys. Lett. 80 3817–19 Collins P G et al 2000 Science 287 1801–4 Derycke V et al 2002 Appl. Phys. Lett. 80 2773–5 Kaminishi D et al 2005 Appl. Phys. Lett. 86 11315 Lin Y-M, Appenzeller T and Avoris P Ph 2004 Nano Lett. 4 947–50 Takeuchi D, Yamanaka S and Okushi H 2002 Diamond Relat. Mater. 11 355–8 Nisihimura K et al 1994 J. Appl. Phys. 76 8142–5 Wang C X et al 2004 Carbon 42 317–21 Nagats M et al 2002 Diamond Relat. Mater. 11 976–9 Itahashi M et al 2001 Diamond Relat. Mater. 10 2118–24 [164] [165] [166] [167] [168] [169] [170] [171] [172] [173] [174] [175] [176] [177] [178] [179] [180] [181] [182] [183] [184] [185] [186] [187] [188] [189] [190] [191] [192] [193] [194] [195] [196] [197] [198] [199] [200] [201] [202] [203] [204] [205] [206] [207] [208] [209] [210] [211] [212] [213] [214] [215] [216] Wu Y-H et al 2002 Diamond Relat. Mater. 11 804–8 Ishizaka H et al 2002 Diamond Relat. Mater. 11 378–81 Eggebert M S et al 2001 Diamond Relat. Mater. 10 744–9 Thumm M 2001 Diamond Relat. Mater. 10 1692–9 Kubovich M et al 2004 Diamond Relat. Mater. 13 802–7 Wind S J et al 2002 Appl. Phys. Lett. 80 431–3 Postma H W Ch et al 2001 Science 293 76–9 Radosavljevic M et al 2002 Nano Lett. 2 761–4 Collins P G, Arnold M S and Avouris P 2001 Science 292 706 Kamo M et al 1984 US Patent 4434188 Aroutiounian V, Martirosyan Kh and Soukiassian P 2004 J. Phys. D: Appl. Phys. 37 L25–8 Watt G A, Newton M E and Baker J M 2001 Diamond Relat. Mater. 10 1681–3 Hayashi Y et al 2002 Diamond Relat. Mater. 11 499–503 Hong S-P et al 2002 Diamond Relat. Mater. 11 877–81 Fujisaki T et al 2002 Diamond Relat. Mater. 11 478–81 Yoshikawa H, Morel C and Koga Y 2001 Diamond Relat. Mater. 10 1588–91 Pleuler E et al 2002 Diamond Relat. Mater. 11 467–71 Gerbi B et al 2003 Appl. Phys. Lett. 83 2001–3 Looi H J et al 1999 Diamond Relat. Mater. 8 966 Werner M M 2003 Semicond. Sci. Technol. 18 S41–46 Prins J F 1982 Appl. Phys. Lett. 41 950 Aleksov A, Denisenko A and Kohn E 2000 Solid-State Electron. 44 369–75 Aleksov A et al 2003 Semicond Sci. Technol. 18 S59–66 Ebert W et al 1994 IEEE Electron. Device Lett. 15 289 Vescan A et al 1997 IEEE Electron. Device Lett. 18 556 Shiomi H et al 1995 IEEE Electron. Device Lett. 16 36 Taniuchi H et al 2001 IEEE Electron Device Lett. 22 390 Aleksov A et al 2002 Proc. 23rd Int. Conf., MIEL vol 1 (Nis, Yugoslavia, 12–15 May) Matsunaga K et al 2000 Int. Electron Devices Meeting, IEDM (San Francisco, CA, 10–13 December) paper 16.6.1 Ebert W et al 1997 Diamond Relat. Mater. 6 329 Chatty K et al 2000 IEEE Electron Device Lett. 21 356 Werner M M 1992 Proc. 3rd Eur. Int. Conf. Diamond and Diamond Like Related Coatings (Germany) Schmid P, Adamchik M and Kohn E 2003 Semicond. Sci. Technol. S72–6 Adamschik M et al 2002 Diamond Relat. Mater. 11 672–6 Stobbe S, Lindelof P E and Nygard J 2006 Semicond. Sci. Technol. 21 S10–16 Milne W I 2003 Semicond. Sci. Technol. 18 S81–5 Nakahata H et al 2003 Semicond. Sci. Technol. 18 S96–104 Wu T-T and Chen Y-Y 2002 Trans. Ultrason. Ferroelectr. Freq. Control 49 311–6 Whaley D R et al 2000 IEEE Trans. Plasma Sci. 28 727 Wisitsora-at A et al 2003 J. Vac. Sci. Technol. B 21 614 Wisitsora-at A et al 2003 J. Vac. Sci. Technol. B 21 1665 Wisitsora-at A et al 2003 J. Vac. Sci. Technol. B 21 1671 Seo S-H, Shin W-Ch and Park J-S 2002 Thin Solid Films 416 190–6 Milne W I et al 2004 Curr. Appl. Phys. 4 513–17 Xu X et al 2002 Thin Solid Films 416 294–301 Monroy E et al 2000 J. Appl. Phys. 88 2081–91 Qiao D et al 2000 J. Appl. Phys. 87 801–4 Ashby C I H et al 2000 Appl. Phys. Lett. 77 3233 Chernyak L, Osinsky A and Schulte A 2001 Solid-State Electron. 45 1687–702 Bachtold A et al 2001 Science 294 1317 Heinze S et al 2000 Science 289 94 Choi W B et al 2003 Appl. Phys. Lett. 82 172–4 Collins P G, Arnold S M and Avouris P 2001 Science 292 706 Radosavljevic M et al 2002 Nano Lett. 2 761–4 Fuhrer M S et al 2002 Nano Lett. 2 755–9 6385 View publication stats