Silicon Carbide Technology:SiC Electronic Devices and Circuits
SiC Electronic Devices and Circuits
This section briefly summarizes a variety of SiC electronic device designs broken down by major application areas. SiC process and material technology issues limiting the capabilities of various SiC device topologies are highlighted as key issues to be addressed in further SiC technology maturation. Throughout this section, it should become apparent to the reader that the most difficult general challenge preventing SiC electronics from fully attaining beneficial capabilities is attaining long-term high operational reli- ability, while operating in previously unattained temperature and power density regimes. Because many device reliability limitations can be traced to fundamental material and junction/interface issues already mentioned in Sections 5.4 and 5.5, efforts to enable useful (i.e., reliable) SiC electronics should focus on improvements to these fundamental areas.
SiC Optoelectronic Devices
The wide bandgap of SiC is useful for realizing short-wavelength blue and ultraviolet (UV) optoelec- tronics. 6H-SiC-based pn junction light-emitting diodes (LEDs) were the first semiconductor devices to cover the blue portion of the visible color spectrum, and became the first SiC-based devices to reach high-volume commercial sales [153]. Because SiC’s bandgap is indirect (i.e., the conduction minimum and valence band maximum do not coincide in crystal momentum space), luminescent recombination is inherently inefficient [154]. Therefore, LEDs based on SiC pn junctions were rendered quite obsolete by the emergence of much brighter, much more efficient direct-bandgap Group III-nitride (III-N such as GaN, and InGaN) blue LEDs [155]. However, SiC wafers are still employed as one of the substrates (along with sapphire) for growth of III-N layers used in high-volume manufacture of green and blue nitride-based LEDs.
SiC has proven much more efficient at absorbing short-wavelength light, which has enabled the realization of SiC UV-sensitive photodiodes that serve as excellent flame sensors in turbine-engine combustion monitoring and control [153,156]. The wide bandgap of 6H-SiC is useful for realizing low photodiode dark currents as well as sensors that are blind to undesired near-infrared wavelengths produced by heat and solar radiation. Commercial SiC-based UV flame sensors, again based on epitaxially grown dry-etch mesa-isolated 6H-SiC pn junction diodes, have successfully reduced harmful pollution emissions from gas-fired ground-based turbines used in electrical power generation systems [156]. The low dark-currents of SiC diodes are also useful for X-ray, heavy ion, and neutron detection in nuclear reactor monitoring and enhanced scientific studies of high-energy particle collisions and cosmic radiation [157,158].
SiC RF Devices
The main use of SiC RF devices appears to lie in high-frequency solid-state high-power amplification at frequencies from around 600 MHz (UHF-band) to perhaps as high as a few gigahertz (X-band). As discussed in far greater detail in References 5, 6, 25, 26, 159, and elsewhere, the high breakdown voltage and high thermal conductivity coupled with high carrier saturation velocity allow SiC RF transistors to handle much higher power densities than their silicon or GaAs RF counterparts, despite SiC’s disadvantage in low-field carrier mobility (Table 5.1). The higher thermal conductivity of SiC is also crucial in minimizing channel self-heating so that phonon scattering does not seriously degrade carrier velocity. These material advantage RF power arguments apply to a variety of different transistor structures such as MESFETs and static induction transistors (SITs) and other wide bandgap semiconductors (such as Group III-nitrides) besides SiC. The high power density of wide bandgap transistors will prove quite useful in realizing solid-state transmitter applications, where higher power with smaller size and mass are crucial. Fewer transistors capable of operating at higher temperatures reduce matching and cooling requirements, leading to reduced overall size and cost of these systems.
SiC-based high-frequency RF MESFETs are now commercially available [40]. However, it is important to note that this occurred after years of fundamental research tracked down and eliminated poor reliability owing to charge-trapping effects arising from immature semi-insulating substrates, device epilayers, and surface passivation [159]. One key material advancement that enabled reliable operation was the devel- opment of “high-purity”semi-insulating SiC substrates (needed to minimize parasitic device capacitances) with far less charge trapping induced than the previously developed vanadium-doped semi-insulating SiC wafers. SiC MESFET devices fabricated on semi-insulating substrates are conceivably less susceptible to adverse yield consequences arising from micropipes than vertical high-power switching devices, primarily because a c-axis micropipe can no longer short together two conducting sides of a high field junction in most areas of the lateral channel MESFET structure.
SiC mixer diodes also show excellent promise for reducing undesired intermodulation interference in RF receivers [160–162]. More than 20 dB dynamic range improvement was demonstrated using non- optimized SiC Schottky diode mixers. Following further development and optimization, SiC-based mixers should improve the interference immunity in situations (such as in aircraft or ships) where receivers and high-power transmitters are closely located.
SiC High-Temperature Signal-Level Devices
Most analog signal conditioning and digital logic circuits are considered “signal level” in that individual transistors in these circuits do not typically require any more than a few milliamperes of current and <20 V to function properly. Commercially available silicon-on-insulator circuits can perform complex digital and analog signal-level functions up to 300°C when high-power output is not required [163]. Besides ICs in which it is advantageous to combine signal-level functions with high-power or unique SiC sensors/MEMS onto a single chip, more expensive SiC circuits solely performing low-power signal-level functions appear largely unjustifiable for low-radiation applications at temperatures below 250–300°C [7].
As of this writing, there are no commercially available semiconductor transistors or integrated circuits (SiC or otherwise) for use in ambient temperatures above 300°C. Even though SiC-based high-temperature laboratory prototypes have improved significantly over the last decade, achieving long-term operational reliability remains the primary challenge of realizing useful 300–600°C devices and circuits. Circuit technologies that have been used to successfully implement VLSI circuits in silicon and GaAs such as CMOS, ECL, BiCMOS, DCFL, etc., are to varying degrees candidates for T > 300°C SiC-integrated circuits. High-temperature gate-insulator reliability (Section 5.5.5) is critical to the successful realization of MOSFET-based integrated circuits. Gate-to-channel Schottky diode leakage limits the peak operating temperature of SiC MESFET circuits to around 400°C (Section 5.5.3.2). Therefore, pn junction-based devices such as bipolar junction transistors (BJTs) and junction field effect transistors (JFETs), appear to be stronger (at least in the nearer term) candidate technologies to attain long-duration operation in 300–600°C ambients. Because signal-level circuits are operated at relatively low electric fields well below the electrical failure voltage of most dislocations, micropipes and other SiC dislocations affect signal- level circuit process yields to a much lesser degree than they affect high-field power device yields.
As of this writing, some discrete transistors and small-scale prototype logic and analog amplifier SiC- based ICs have been demonstrated in the laboratory using SiC variations of NMOS, CMOS, JFET, and MESFET device topologies [164–170]. However, none of these prototypes are commercially viable as of this writing, largely owing to their inability to offer prolonged-duration electrically stable operation at ambient temperatures beyond the ~250–300°C realm of silicon-on-insulator technology. As discussed in Section 5.5, a common obstacle to all high-temperature SiC device technologies is reliable long-term operation of contacts, interconnect, passivation, and packaging at T > 300°C. By incorporating highly durable high-temperature ohmic contacts and packaging, prolonged continuous electrical oper- ation of a packaged 6H-SiC field effect transistor at 500°C in oxidizing air environment was recently demonstrated [111,112,149].
As further improvements to fundamental SiC device processing technologies (Section 5.5) are made, increasingly durable T > 300°C SiC-based transistor technology will evolve for beneficial use in harsh- environment applications. Increasingly complex high-temperature functionality will require robust circuit designs that accommodate large changes in device operating parameters over the much wider temperature ranges (as large as 650°C spread) enabled by SiC. Circuit models need to account for the fact that SiC device epilayers are significantly “frozen-out” owing to deeper donor and acceptor dopant ionization energies, so that nontrivial percentages of device-layer dopants are not ionized to conduct current near room temperature [171]. Because of these carrier freeze-out effects, it will be difficult to realize SiC-based ICs operational at junction temperatures much lower than –55°C (the lower end of U.S. Mil-Spec. temperature range).
SiC High-Power Switching Devices
The inherent material properties and basic physics behind the large theoretical benefits of SiC over silicon for power switching devices were discussed Section 5.3.2. Similarly, it was discussed in Section 5.4.5 that crystallographic defects found in SiC wafers and epilayers are presently a primary factor limiting the commercialization of useful SiC high-power switching devices. This section focuses on the additional developmental aspects of SiC power rectifiers and power switching transistor technologies.
Most SiC power device prototypes employ similar topologies and features as their silicon-based counterparts such as vertical flow of high current through the substrate to maximize device current using minimal wafer area (i.e., maximize current density) [18]. In contrast to silicon, however, the relatively low conductivity of present-day p-type SiC substrates (Section 5.4.3) dictates that all vertical SiC power device structures be implemented using n-type substrates in order to achieve beneficially high vertical current densities. Many of the device design trade-offs roughly parallel well-known silicon power device trade-offs, except for the fact that numbers for current densities, voltages, power densities, and switching speeds are much higher in SiC.
For power devices to successfully function at high voltages, peripheral breakdown owing to edge- related electric field crowding [15,18,104] must be avoided through careful device design and proper choice of insulating/passivating dielectric materials. The peak voltage of many prototype high-voltage SiC devices has often been limited by destructive edge-related breakdown, especially in SiC devices capable of blocking multiple kilovolts. In addition, most testing of many prototype multikilovolt SiC devices has required the device to be immersed in specialized high-dielectric strength fluids or gas atmospheres to minimize damaging electrical arcing and surface flashover at device peripheries. A variety of edge-termination methodologies, many of which were originally pioneered in silicon high- voltage devices, have been applied to prototype SiC power devices with varying degrees of success, including tailored dopant and metal guard rings [172–179]. The higher voltages and higher local electric fields of SiC power devices will place larger stresses on packaging and on wafer insulating materials, so some of the materials used to insulate/passivate silicon high-voltage devices may not prove sufficient for reliable use in SiC high-voltage devices, especially if those devices are to be operated at high temperatures.
SiC High-Power Rectifiers
The high-power diode rectifier is a critical building block of power conversion circuits. Recent reviews of experimental SiC rectifier results are given in References 3, 134, 172, 180, and 181. Most important SiC diode rectifier device design trade-offs roughly parallel well-known silicon rectifier trade-offs, except for the fact that current densities, voltages, power densities, and switching speeds are much higher in SiC. For example, semiconductor Schottky diode rectifiers are majority carrier devices that are well known to exhibit very fast switching owing to the absence of minority carrier charge storage that dominates (i.e., slows, adversely resulting in undesired waste power and heat) the switching operation of bipolar pn junction rectifiers. However, the high breakdown field and wide energy bandgap permit operation of SiC metal–semiconductor Schottky diodes at much higher voltages (above 1 kV) than is practical with silicon- based Schottky diodes that are limited to operation below ~200 V owing to much higher reverse-bias thermionic leakage.
SiC Schottky Power Rectifiers.
4H-SiC power Schottky diodes (with rated blocking voltages up to 1200 V and rated on-state currents up to 20 A as of this writing) are now commercially available [40,113]. The basic structure of these unipolar diodes is a patterned metal Schottky anode contact residing on top of a relatively thin (roughly of the order of 10 µm in thickness) lightly n-doped homoepitaxial layer grown on a much thicker (around 200–300 µm) low-resistivity n-type 4H-SiC substrate (8° off axis, as discussed in Section 5.4.4.2) with backside cathode contact metallization [172,182]. Guard ring structures (usually p-type implants) are usually employed to minimize electric field crowding effects around the edges of the anode contact. Die passivation and packaging help prevent arcing/surface flashover harmful to reliable device operation.
The primary application of these devices to date has been switched-mode power supplies, where (consistent with the discussion in Section 5.3.2) the SiC Schottky rectifier’s faster switching with less power loss has enabled higher frequency operation and shrinking of capacitors, inductors and the overall power supply size and weight [3,23]. In particular, the effective absence of minority carrier charge storage enables the unipolar SiC Schottky devices to turn off much faster than the silicon rectifiers (which must be pn junction diodes above ~200 V blocking) which must dissipate injected minority carrier charge energy when turned off. Even though the part cost of SiC rectifiers has been higher than competing silicon rectifiers, an overall lower power supply system cost with useful per- formance benefits is nevertheless achieved. It should be noted, however, that changes in circuit design are sometimes necessary to best enhance circuit capabilities with acceptable reliability when replacing silicon with SiC components.
As discussed in Section 5.4.5, SiC material quality presently limits the current and voltage ratings of SiC Schottky diodes. Under high forward bias, Schottky diode current conduction is primarily limited by the series resistance of the lightly doped blocking layer. The fact that this series resistance increases with temper- ature (owing to decreased epilayer carrier mobility) drives equilization of high forward currents through each diode when multiple Schottky diodes are paralleled to handle higher on-state current ratings [17].
Bipolar and Hybrid Power Rectifiers.
For higher voltage applications, bipolar minority carrier charge injection (i.e., conductivity modulation) should enable SiC pn diodes to carry higher current densities than unipolar Schottky diodes whose drift regions conduct solely using dopant-atom majority carriers [19–21,172,180]. Consistent with silicon rectifier experience, SiC pn junction generation-related reverse leakage is usually smaller than thermionic- assisted Schottky diode reverse leakage. As with silicon bipolar devices, reproducible control of minority carrier lifetime will be essential in optimizing the switching-speed versus on-state current density per- formance trade-offs of SiC bipolar devices for specific applications. Carrier lifetime reduction via inten- tional impurity incorporation and introduction of radiation-induced defects appears feasible. However,
As of this writing, SiC bipolar power rectifiers are not yet commercially available. Poor electrical reliability caused by electrically driven expansion of 4H-SiC epitaxial layer stacking faults initiated from basal plane dislocation defects (Table 5.2) effectively prevented concerted efforts for commer- cialization of 4H-SiC pn junction diodes in the late 1990s [63,74,184]. In particular, bipolar electron–hole recombination that occurs in forward-biased pn junctions drove the enlargement of stacking disorder in the 4H-SiC blocking layer, forming an enlarging quantum well (based on narrower 3C-SiC bandgap) that effectively degrades transport (diffusion) of minority carriers across the lightly doped junction blocking layer. As a result, the forward voltages of 4H-SiC pn rectifiers required to maintain rated on-state current increase unpredictably and undesirably over time. As discussed in Section 5.4.5, research toward understanding and overcoming this material defect-induced problem has made important progress, so that hopefully SiC bipolar power devices might become commercialized within a few years [39,41].
A drawback of the wide bandgap of SiC is that it requires larger forward-bias voltages to reach the turn-on “knee” of a diode where significant on-state current begins flowing. In turn, the higher knee voltage can lead to an undesirable increase in on-state power dissipation. However, the benefits of 100× decreased drift region resistance and much faster dynamic switching should greatly overcome SiC on- state knee voltage disadvantages in most high-power applications. While the initial turn-on knee of SiC pn junctions is higher (around 3 V) than for SiC Schottky junctions (around 1 V), conductivity modulation enables SiC pn junctions to achieve lower forward voltage drop for higher blocking voltage applications [172,180].
Hybrid Schottky/pn rectifier structures first developed in silicon that combine pn junction reverse blocking with low Schottky forward turn-on should prove extremely useful in realizing application- optimized SiC rectifiers [134,172,180,181]. Similarly, combinations of dual Schottky metal structures and trench pinch rectifier structures can also be used to optimize SiC rectifier forward turn-on and reverse leakage properties [185].
SiC High-Power Switching Transistors
Three terminal power switches that use small drive signals to control large voltages and currents (i.e., power transistors) are also critical building blocks of high-power conversion circuits. However, as of this writing, SiC high-power switching transistors are not yet commercially available for beneficial use in power system circuits. As well summarized in References 134, 135, 172, 180, and 186–188, a variety of improving three-terminal SiC power switches have been prototyped in recent years.
The present lack of commercial SiC power switching transistors is largely due to several technological difficulties discussed elsewhere in this chapter. For example, all high-power semiconductor transistors contain high-field junctions responsible for blocking current flow in the off-state. Therefore, performance limitations imposed by SiC crystal defects on diode rectifiers (Sections 5.4.5 and 5.6.4.1) also apply to SiC high-power transistors. Also, the performance and reliability of inversion channel SiC-based MOS field- effect gates (i.e., MOSFETs, IGBTs, etc.) has been limited by poor inversion channel mobilities and questionable gate-insulator reliability discussed in Section 5.5.5. To avoid these problems, SiC device structures that do not rely on high-quality gate insulators, such as the MESFET, JFET, BJT, and depletion-channel MOSFET, have been prototyped toward use as power switching transistors. However, these other device topologies impose non-standard requirements on power system circuit design that make them unattractive compared with the silicon-based inversion-channel MOSFETs and IGBTs. In particular, silicon power MOSFETs and IGBTs are extremely popular in power circuits largely because their MOS gate drives are well insulated from the conducting power channel, require little drive signal power, and the devices are “normally off ” in that there is no current flow when the gate is unbiased at 0 V. The fact that the other device topologies lack one or more of these highly circuit-friendly aspects has contributed to the inability of SiC-based devices to beneficially replace silicon-based MOSFETs and IGBTs in power system applications. As discussed in Section 5.5.5, continued substantial improvements in 4H-SiC MOSFET technology will hopefully soon lead to the commercialization of 4H-SiC MOSFETs. In the meantime, advantageous high- voltage switching by pairing a high-voltage SiC JFET with a lower-voltage silicon power MOSFETs into a single module package appears to be nearing practical commercialization [188]. Numerous designs for SiC doped-channel FETs (with both lateral and vertical channels) have been prototyped, including depletion- channel (i.e., buried or doped channel) MOSFETs, JFETs, and MESFETs [187]. Even though some of these have been designed to be “normally-off ” at zero applied gate bias, the operational characteristics of these devices have not (as of this writing) offered sufficient benefits relative to cost to enable commercialization.
Substantial improvements to the gain of prototype 4H-SiC power BJTs have been achieved recently, in large part by changing device design to accommodate for undesired large minority carrier recombi- nation occurring at p-implanted base contact regions [103]. IGBTs, thyristors, Darlington pairs, and other bipolar power device derivatives from silicon have also been prototyped in SiC [134,180,186]. Optical transistor triggering, a technique quite useful in previous high-power silicon device applications, has also been demonstrated for SiC bipolar devices [189]. However, because all bipolar power transistors operate with at least one pn junction injecting minority carriers under forward bias, crystal defect-induced bipolar degradation discussed for pn junction rectifiers (Section 5.6.4.1.2) also applies to the performance of bipolar transistors. Therefore, the effective elimination of basal plane dislocations from 4H-SiC epil- ayers must be accomplished before any power SiC bipolar transistor devices can become sufficiently reliable for commercialization. SiC MOS oxide problems (Section 5.5.5) will also have to be solved to realize beneficial SiC high-voltage IGBTs. However, relatively poor p-type SiC substrate conductivity may force development of p-IGBTs instead of n-IGBT structures that presently dominate in silicon technology. As various fundamental SiC power device technology challenges are overcome, a broader array of SiC power transistors tackling increasingly widening voltage, current, and switching speed specification will enable beneficial new power system circuits.
SiC MicroElectromechanical Systems (MEMS) and Sensors
As described in Hesketh’s chapter on micromachining in this book, the development and use of silicon- based MEMS continues to expand. While the previous sections of this chapter have centered on the use of SiC for traditional semiconductor electronic devices, SiC is also expected to play a significant role in emerging MEMS applications [124,190]. SiC has excellent mechanical properties that address some shortcomings of silicon-based MEMS such as extreme hardness and low friction reducing mechanical wear-out as well as excellent chemical inertness to corrosive atmospheres. For example, SiCs excellent durability is being examined as enabling for long-duration operation of electric micromotors and micro jet-engine power generation sources where the mechanical properties of silicon appear to be insufficient [191].
Unfortunately, the same properties that make SiC more durable than silicon also make SiC more difficult to micromachine. Approaches to fabricating harsh-environment MEMS structures in SiC and prototype SiC-MEMS results obtained to date are reviewed in References 124 and 190. The inability to perform fine-patterned etching of single-crystal 4H- and 6H-SiC with wet chemicals (Section 5.5.4) makes micromachining of this electronic-grade SiC more difficult. Therefore, the majority of SiC micro- machining to date has been implemented in electrically inferior heteroepitaxial 3C-SiC and polycrystalline SiC deposited on silicon wafers. Variations of bulk micromachining, surface micromachining, and micro- molding techniques have been used to fabricate a wide variety of micromechanical structures, including resonators and micromotors. A standardized SiC on silicon wafer micromechanical fabrication process foundry service, which enables users to realize their own application-specific SiC micromachined devices while sharing wafer space and cost with other users, is commercially available [192].
For applications requiring high temperature, low-leakage SiC electronics not possible with SiC layers deposited on silicon (including high-temperature transistors, as discussed in Section 5.6.2), concepts for integrating much more capable electronics with MEMS on 4H/6H SiC wafers with epilayers have also been proposed. For example, pressure sensors being developed for use in higher temperature regions of jet engines are implemented in 6H-SiC, largely owing to the fact that low junction leakage is required to achieve proper sensor operation [152,193]. On-chip 4H/6H integrated transistor electronics that bene- ficially enable signal conditioning at the high-temperature sensing site are also being developed [112]. With all micromechanical-based sensors, it is vital to package the sensor in a manner that minimizes the imposition of thermomechanical induced stresses (which arise owing to thermal expansion coefficient mismatches over much larger temperature spans enabled by SiC) onto the sensing elements. Therefore (as mentioned previously in Section 5.5.6), advanced packaging is almost as critical as the use of SiC toward usefully expanding the operational envelope of MEMS in harsh environments.
As discussed in Section 5.3.1, a primary application of SiC harsh-environment sensors is to enable active monitoring and control of combustion engine systems to improve fuel efficiency while reducing pollution. Toward this end, SiC’s high-temperature capabilities have enabled the realization of catalytic metal–SiC and metal-insulator–SiC prototype gas sensor structures with great promise for emission monitoring applications and fuel system leak detection [194,195]. High-temperature operation of these structures, not possible with silicon, enables rapid detection of changes in hydrogen and hydrocarbon content to sensitivities of parts per million in very small-sized sensors that could easily be placed unobtrusively on an engine without the need for cooling. However, further improvements to the reli- ability, reproducibility, and cost of SiC-based gas sensors are needed before these systems will be ready for widespread use in consumer automobiles and aircraft. In general, the same can be said for most SiC MEMS, which will not achieve widespread beneficial system insertion until high reliability in harsh environments is assured via further technology development.
Comments
Post a Comment