Power Semiconductor Devices for Hybrid, Electric, and ... - IEEE Xplore

12 downloads 0 Views 1MB Size Report
ABSTRACT | Power semiconductor devices are key compo- nents in all power electronic systems, particularly in hybrid, electric, and fuel cell vehicles.
INVITED PAPER

Power Semiconductor Devices for Hybrid, Electric, and Fuel Cell Vehicles Insulated Gate Bipolar Transistors (IGBTs), freewheeling diodes and wide-bandgap devices, able to meet vehicle power train conversion needs, are in development. By Z. John Shen, Senior Member IEEE, and Ichiro Omura, Member IEEE

ABSTRACT

|

Power semiconductor devices are key compo-

nents in all power electronic systems, particularly in hybrid, electric, and fuel cell vehicles. This paper reviews the system requirement and latest development of power semiconductor devices including IGBTs, freewheeling diodes, and advanced power module technology in relating to electric vehicle applications. State-of-the-art silicon device technologies, their future trends, and theoretical limits are discussed. Emerging wide bandgap semiconductor devices such as SiC devices and their potential applications in electric vehicles are also reviewed.

KEYWORDS | Diodes; insulated gate bipolar transistors (IGBTs); power modules; power semiconductor devices

I. INTRODUCTION Power semiconductor devices are key components in all power electronic systems, particularly in hybrid, electric, and fuel cell vehicles. High-voltage power devices such as insulated gate bipolar transistors (IGBTs) and freewheeling diodes play a critical role in hybrid electric traction inverters, voltage boost dc/dc converters, fuel cell air compressor motor drives, and other on-aboard power management converters. In addition, low-voltage power MOSFETs and power integrated circuits are widely used in engine control, vehicle dynamic control, vehicle safety,

Manuscript received July 14, 2006; revised September 27, 2006. This work was supported in part by the U.S. National Science Foundation under Award ECS-0454835 and in part by the Ford Motor Company under a Ford University Research Program (URP) grant. Z. J. Shen is with the School of Electrical Engineering and Computer Science, University of Central Florida, Orlando, FL 32816-2450 USA (e-mail: [email protected]). I. Omura is with the Advanced Power Semiconductor Development Group, Toshiba Corporation, Kawasaki 212-8583, Japan (e-mail: [email protected]). Digital Object Identifier: 10.1109/JPROC.2006.890118

778

Proceedings of the IEEE | Vol. 95, No. 4, April 2007

and body electronics subsystems in both electric and conventional internal combustion engine (ICE) vehicles. Power semiconductor devices often dictate the efficiency, cost, and size of the vehicle systems. It is estimated that the power semiconductor chips needed in a typical hybrid electric vehicle may add up to 30% of a 6-in silicon wafer [1]. Power semiconductor devices count for roughly onethird of the total cost of vehicle power electronics, which has a targeted goal of U.S. $7/kW for a volume of 100 000 units set by the automotive industry [2]. The emerging hybrid electric vehicle market presents a tremendous business opportunity for the semiconductor industry but at the same time a great technical challenge in improving performance, reliability, and manufacturing cost of the power semiconductor products. This paper reviews the system requirement and latest development of power semiconductor devices including IGBTs, freewheeling diodes, and advanced power module technology relating to electric vehicle applications. Stateof-the-art silicon device technologies, their future trends, and theoretical limits are discussed. Emerging wide bandgap semiconductor devices such as SiC devices and their potential applications in electric vehicles are also reviewed. The paper is focused on power devices with a voltage rating of 600–1200 V for propulsion applications. Low-voltage power devices with a voltage rating below 100 V for nonpropulsion automotive applications are therefore not discussed in this paper.

I I. SYSTEM RE QUIREMENTS Fig. 1 conceptually illustrates a hybrid powertrain example based on Toyota’s Prius hybrid electric vehicle [1], [3]–[5]. The hybrid parallel and series powertrain consists of an ICE as the primary power source, a motor and motor drive inverter for propulsion, a generator and generator inverter 0018-9219/$25.00 Ó 2007 IEEE

Shen and Omura: Power Semiconductor Devices for Hybrid, Electric, and Fuel Cell Vehicles

Fig. 1. Schematic diagram of hybrid electric vehicle powertrain [1].

for electric power generation, a battery for energy storage, and a voltage boost dc/dc converter. The Toyota Prius hybrid system employs a power split mechanism in which the ICE, motor, and generator are coupled to a planetary gear system to achieve optimum power efficiency. Two IGBT inverters are used to control the motor and generator independently for a wide speed ratio transmission and optimum electric energy flow to and from the battery. The variable voltage boost converter made of IGBTs and freewheeling diodes provides a bus voltage significantly higher than the battery voltage to achieve a higher power output from the propulsion motor. The output power of the inverters for the motor and generator are 50–80 and 30–50 kW, respectively. The output power of the boost converter is roughly 20–30 kW. The hybrid electric vehicle (HEV) system efficiently uses the ICE power to drive the wheels through the power split system (the planetary gear system) so that the actual vehicle power is higher than the motor inverter power. A number of IGBTs and freewheeling diodes with a voltage rating of 850 V or higher and a current rating of several hundred amperes are used in this hybrid powertrain system. The use of high-voltage IGBTs and freewheeling diodes in other hybrid, electric, and fuel cell vehicles is similar to the example described above. However, the number and power rating of the needed power devices vary with specific system configuration and requirements. For example, a single combined motor/generator IGBT inverter is typically used in a series hybrid powertrain, resulting in a

reduced number of IGBTs and freewheeling diodes. But the power rating of such an IGBT inverter is considerably higher than that of the IGBT inverters used in a parallel hybrid powertrain with the same power rating. In fuel cell vehicles, in addition to the propulsion inverter and voltage boost converter, some auxiliary power electronic circuits are needed to control the air compressor and the cooling subsystem of the fuel cell stack. Table 1 lists the system requirement of power semiconductor devices for various types of hybrid, electric, and fuel cell vehicle subsystems. The current ratings of a power semiconductor are mainly related to the energy dissipation and the junction temperature in the device. The maximum continuous current is usually defined as the current that the device is capable of conducting continuously without exceeding the maximum junction temperature. Even though the dc bus voltage of most hybrid electric vehicles is between 200 and 500 V, the voltage ratings of power semiconductor devices are considerably higher than that. This is because the voltage rating of automotive power electronics is mainly determined by the survivability of these devices to the commonly encountered over-voltage transients in the automotive environment, instead of just the maximum operating voltage. It is imperative for the inverters or converters to have a small physical size and weight so that they can fit into the limited space available in the vehicle. The allowable physical volume for the power inverters is about 9.5 l [6] regardless of their actual power ratings. The power density of Vol. 95, No. 4, April 2007 | Proceedings of the IEEE

779

Shen and Omura: Power Semiconductor Devices for Hybrid, Electric, and Fuel Cell Vehicles

Table 1 System Requirement of Power Semiconductor Devices in Hybrid, Electric, and Fuel Cell Vehicles

a 100-kW inverter can therefore be as high as 10 kW/l, which is considerably higher than the motor drive inverters for industrial applications. To achieve such a highpower density, semiconductor power devices are required to increase current density with reducing power losses and to improve chip cooling efficiency against increased heat density on the chips.

III . ACTIVE POWER S WITCH: IGBT As shown in Table 1, IGBTs are the active switching device of choice in nearly all HEV inverters and converters due to their superior current conduction capability over highvoltage power MOSFETs. The IGBT is a switching transistor controlled by a voltage applied to its gate terminal. Device operation and structure are similar to that of a power MOSFET. The principal difference is that the IGBT relies on conductivity modulation to reduce on-state conduction losses. The IGBT has high input impedance and fast turn-on speed like a MOSFET but exhibits an onstate voltage drop and current-carrying capability comparable to that of a bipolar transistor while switching much faster. IGBTs have a clear advantage over MOSFETs in high-voltage applications where conduction losses must be minimized. Since the initial introduction of the IGBT into market in the mid 1980s, the semiconductor industry has made great technological advancement in improving device performance and reducing fabrication cost. Fig. 2 shows the historical trend of the ever-increasing current handling capability of the IGBT with the major technological breakthroughs labeled [7]. The current handling capability 780

Proceedings of the IEEE | Vol. 95, No. 4, April 2007

per unit chip area or the current density of the IGBT has increased by more than 100% since 1990, resulting in more than a 50% reduction in silicon real estate for the same power rating. The improvement has been achieved mainly by optimizing planar device geometry, introducing trench gate structures, and optimizing vertical IGBT structures including the N-base, N-buffer, and P-emitter structures. The IGBT device with latch-up immunity first appeared in the market in mid 1980s [8], [9]. The IGBT employs a planar DMOS structure for the top surface of the IGBT chip with a punch-through (PT) vertical bulk structure. Alternatively, a nonpunch-through (NPT) IGBT structure was

Fig. 2. Evolution of IGBT technology over the last 20 years.

Shen and Omura: Power Semiconductor Devices for Hybrid, Electric, and Fuel Cell Vehicles

introduced shortly after [10]. Intelligent power modules (IPM) were later developed to integrate gate drive and protection circuitry with the main IGBTs into a single power module package [11]. Trench technology was first introduced in low-voltage power MOSFETs in the early 1990s, leading to a dramatic reduction in MOSFET on-resistance. Trench gate structures were subsequently adopted in the IGBT to reduce the MOS channel resistance and the socalled JFET resistance. Fig. 3 illustrates the impact of trench gate technology on IGBT conduction loss reduction. It was found that the IGBT structure with a deep trench gate and relatively wide cell pitch considerably enhances the electron injection efficiency at the emitter side by minimizing the back injection of holes into the P-base. The large cell pitch in this type of IGBT leads to a slight penalty in the MOS channel resistance but is more than compensated by the significant reduction of the N-base conduction resistance. This is especially true for high-voltage IGBTs [12], [13]. Alternatively, an N-type layer can be added under the P-base to block the hole back injection current and enhance electron injection [14], [15]. These two carrier enhancement approaches are widely adopted in the state-of-the-art IGBTs as shown in Fig. 3, providing an optimum stored excess carrier profile similar to that of an ideal PiN diode and hence a reduction in forward voltage. In addition to the trench gate and carrier enhancement techniques, thin wafer technology provides dramatic im-

provement in IGBT performance. Historically, punchthrough IGBT (PT-IGBT) structures have been commonly used ever since the first appearance of IGBT products. As shown in Fig. 4, a PT-IGBT is fabricated on an epitaxial wafer. A moderately doped N-buffer and a lightly doped N-base epitaxial layer are formed on a P+ substrate which serves as the collector of the IGBT (or the emitter of the internal PNP transistor). A PT-IGBT features a fairly thin N-base for improved forward conduction characteristics. When the IGBT is in a voltage blocking state, the space charge in the N-buffer layer keeps the depletion region of the main PN junction from reaching through the P+ substrate. The PT-IGBT realizes low on-state conduction loss with the short N-base and high level hole injection from the P+ substrate. The large hole injection efficiency of the P+ substrate, however, causes large amounts of the stored minority carriers near the P+ substrate, which results in a long tail current and large switching power loss. Therefore, the turn-off time of PT-IGBTs must be controlled by reducing the stored carrier with a carrier recombination process. For this purpose, the use of carrier lifetime control techniques is indispensable for PT-IGBTs. Carrier lifetime control techniques help minimize the switching power loss but inadvertently increase the conduction loss with the reduction of stored excess carriers in the entire N-base. Furthermore, the reduced carrier lifetime

Fig. 3. Impact of trench gate technology on IGBT conduction loss reduction.

Vol. 95, No. 4, April 2007 | Proceedings of the IEEE

781

Shen and Omura: Power Semiconductor Devices for Hybrid, Electric, and Fuel Cell Vehicles

Fig. 4. Thin wafer technology satisfies both minimum N-base thickness and optimum hole injection from backside P-emitter, realizing low conduction loss and switching loss.

demonstrates a temperature dependence which may impose an undesirable negative thermal coefficient on some PT-IGBTs. The nonpunch-through IGBTs (NPT-IGBTs) were proposed a long time ago as an alternative approach to the epitaxial-based PT-IGBT [10]. An NPT-IGBT features a relatively long N-base without the N-buffer layer. The NPT-IGBT is fabricated on a floating zone (FZ) silicon wafer without using any epitaxial layers, resulting in a significant cost saving in starting wafers. The advantage of NPT-IGBTs is due to the fact that the P-emitter of the internal PNP transistor is formed by a backside ion implantation process so that the hole injection efficiency of the P-emitter can be controlled by its doping profile. The P-emitter of the NPT-IGBT has a junction depth typically less than 1 m and a reasonably low doping concentration to ensure a sufficiently low hole injection efficiency. No carrier lifetime control technique is required for NPT-IGBTs because the stored carriers are limited by the shallow and light P-emitter. The combination of the shallow P-emitter and the long carrier lifetime in the N-base offers a significant advantage in the turn-off 782

Proceedings of the IEEE | Vol. 95, No. 4, April 2007

characteristics and thermal coefficient of the IGBT. The conduction loss, on the other hand, is relatively high due to the long N-base. Thin wafer punch-through IGBTs represents the latest development in IGBT technology which features a short N-base and an N-buffer layer. The difference between this new type of IGBT and the conventional PT-IGBT is that the former uses a shallow P-emitter and no carrier lifetime killing techniques very much like the conventional NPT-IGBT. The concept of the thin wafer nonpunchthrough structure was first demonstrated in the gate commutated thyristor (GCT) and high-voltage IGBTs with an N-base width of 400 m. However, the ideal wafer thickness for IGBTs of 600–1200 V is below 100 m which mandates a dedicated thin wafer fabrication process and presents a great challenge in wafer manufacturing. Thin wafer process breakthrough was first reported in 1996 and it has since triggered the development effort of a 600–1200-V thin wafer PT-IGBT [16], [17]. The thin wafer PT-IGBT shows significant improvement over the conventional IGBTs because it offers the combined performance benefits of both PT and NPT IGBTs in terms of

Shen and Omura: Power Semiconductor Devices for Hybrid, Electric, and Fuel Cell Vehicles

Table 2 Comparison of Conventional PT-IGBT, NPT-IGBT and Thin Wafer PT_IGBT

BnanoIGBT[ structures, the device concept will definitely become more practical with the maturity of nanoelectronics fabrication technology.

IV. FREEWHEELING DIODES

conduction loss, turn-off speed, and thermal coefficients [18]–[20]. Fig. 4 illustrates the device structure, electrical field at the blocking state, and stored carrier profile during the on state of the PT-IGBT, NPT-IGBT, and thin wafer PT-IGBT. Table 2 provides a detailed comparison of these three types of IGBT technology. Obviously, the thin wafer PT-IGBT represents the best technical solution for the hybrid, electric, and fuel cell vehicle applications. As impressive as the technological advancement in IGBT technology has been during the past two decades, further improvement in IGBT performance and cost effectiveness is still possible in the future. It was recently proposed that the theoretical limit of IGBT current handling capability is several times higher than that of the state-of-the-art IGBTs [21]. For example, the theoretical forward voltage drop of a 600-V ideal IGBT can be as low as 1 V at a current density over 500 A/cm2 , when comparing with 1.4 V at a current density of 100 A/cm2 for a typical IGBT product available today. It is worth noting that such a current handling capability is far superior not only to that of the state-of-the-art silicon IGBT or superjunction power MOSFETs but also to that of the best performing SiC and GaN wide bandgap semiconductor devices reported recently. It will be possible to operate an IGBT at a current density as high as 300 A/cm2 in the future by further optimizing device design and developing an ultra-low thermal resistance package. This will result in a significant reduction in the size and cost of IGBT chips and those of the HEV inverters and converters where these IGBTs are used. One suggested approach to achieve such an ideal IGBT performance is to realize a nanometer-scale trench gate structure of 40–100 nm in width to further reduce the MOS channel resistance and enhance electron injection to achieve a more ideal stored carrier profile in the N-base [21], [22]. While many technical and economical challenges remain to be solved in realizing such

Diodes are the simplest semiconductor device, yet have a major impact on the overall efficiency of HEV inverters and converters. They are mainly used as the anti-parallel freewheeling diodes to conduct the load current during IGBT turn-off. In addition to the voltage rating, current rating, and forward voltage drop, the recovery characteristics of freewheeling diodes dictate the selection of rectifiers for fast switching power circuits. A power diode requires a finite time to switch from off state (reverse bias) to on state (forward bias) and vice versa. Both the recovery times and the shapes of the waveforms are affected by the intrinsic properties of the diode and by the external circuit. The recovery process from on to off is termed Breverse recovery.[ In hard switching applications, IGBT turn-on losses increase as the reverse recovery rises and recovery time is prolonged. To minimize the turn-on losses, the freewheeling diodes need to have both fast and soft recovery characteristics. A diode with a snappy reverse recovery characteristic and hence high recovery di=dt is problematic with transient voltages. During fabrication of the freewheeling diodes, special care must be taken to reduce the reverse recovery charge and recovery time by decreasing the doping concentration and diffusion depth of the anode P region. Fig. 5 compares the reverse recovery characteristics of a slow and snappy diode with a fast and soft recovery diode. The slow and snappy diode demonstrates a larger peak recovery current yet a higher di=dt during the recovery phase. In contrast, the fast and soft diode shows a reduced peak recovery current and di=dt at the same time. Ideally, the freewheeling diodes used in HEV inverters/converters should have the following features [23]: 1) low forward voltage and positive temperature coefficient for safe parallel diode operation; 2) low reverse recovery losses, soft recovery, and ruggedness against dynamic avalanching; 3) stable reverse blocking capability with low leakage current at high temperatures; and 4) surge current capability, avalanche withstand capability, and a low forward recovery overshoot voltage during the diode turn-on transient period. Several techniques are employed in the design of a PNN+ diode to improve the reverse recovery characteristics of freewheeling diodes, including PN junction injection efficiency control, local carrier lifetime control, and deep diffusion control [23]. The PN junction injection control technique is based on the use of a low PN junction injection efficiency to control the gradient of the excess carrier concentrations. The shape and distribution of the stored charge in the N-drift region influences the reverse recovery characteristics of the diode. An increasing carrier distribution toward the NN+ junction is preferred for Vol. 95, No. 4, April 2007 | Proceedings of the IEEE

783

Shen and Omura: Power Semiconductor Devices for Hybrid, Electric, and Fuel Cell Vehicles

achieving a smaller reverse recovery charge and softer recovery characteristics. The low PN injection efficiency also aids in achieving positive temperature coefficient on-state characteristics, which are necessary for parallel operation of diodes. Local carrier lifetime control processes such as proton or helium implantation are used for controlling the axial carrier lifetime profile. This technique creates a region of recombination centers near the PN junction to effectively reduce the reverse recovery charge and maintain a soft recovery characteristic. On the other hand, the region near the NN+ junction maintains a high carrier lifetime and provides the additional residual charge for soft recovery. A uniform lifetime control process using electron irradiation is often used in addition to local lifetime control to improve the softness of the reverse recovery characteristics of the diode. Conventional punch-through epitaxial PNN+ diode structures usually demonstrate snappy recovery characteristics because of the sudden disappearance of the minority carriers stored in the N drift region when the depletion layer reaches the N+ region. Snappy recovery can be prevented by adding an N-buffer layer with higher doping levels between the N-drift and N+ regions. The N-buffer region prevents the spreading out of the depletion layer, providing the residual charge needed for soft recovery. The doping concentration of the N-buffer region should be high enough to stop the depletion layer but low enough to maintain necessary conductivity modulation. An alterna-

Fig. 5. Reverse recovery characteristics of freewheeling diodes.

784

Proceedings of the IEEE | Vol. 95, No. 4, April 2007

tive approach to the epitaxial device structure is to use a thin float-zone wafer with a controlled deep diffused N+ layer together with the controlled graded stored charge, leading to a soft punch-through near the NN+ junction. The use of thin wafers results in reduction of both forward voltage drop and cost of starting wafer materials, similar to the latest development of thin wafer PT-IGBT technology.

V. Si C AND OTHE R WI DE BANDGAP SEMICONDUCTOR DEVICES In the past two decades, the technological advancement has pushed the performance of state of the art power semiconductor devices to the theoretical limit of silicon material. Silicon carbide (SiC), gallium nitride (GaN), and other wide bandgap (WBG) semiconductor devices have become the focus of the power semiconductor research community. SiC is generally considered the most promising semiconductor material to replace silicon in future power electronic systems. SiC or other WBG power devices offer several benefits over their silicon counterparts. First, the higher breakdown electric field strength of SiC allows a much thinner drift region and thus a much smaller specific on-resistance of SiC devices than their silicon counterparts. Secondly, the low on-resistance of SiC devices for a voltage rating of 600–2000 V allows the use of majority-carrier devices like MOSFET and Schottky diodes rather than minority-carrier devices such as IGBT

Shen and Omura: Power Semiconductor Devices for Hybrid, Electric, and Fuel Cell Vehicles

and PiN diodes. This results in a much reduced switching loss absence of the charge storage effect. Lower switching losses will further allow higher switching frequency and subsequently smaller and less expensive passive components such as filter inductors and capacitors. Thirdly, the larger bandgap results in a lower intrinsic carrier concentration and higher operating junction temperature. In principle, SiC devices could operate at a junction temperature as high as 500  C, as compared to a typical 150  C maximum junction temperature of silicon devices (a small percentage of silicon products are rated up to 200  C). The increased operating temperature will reduce the weight, volume, cost, and complexity of thermal management systems. Lastly, the very high thermal conductivity of SiC reduces the thermal resistance of the device die. Significant progress has been made in research and development of SiC and other WBG device technologies during the past decade. Various types of SiC switching devices and diodes have been developed and reported. The 600–1200-V SiC Schottky diodes are commercially available, which can replace Si freewheeling diodes in HEV inverters. SiC Schottky diodes have demonstrated superior performance to that of similar silicon PN diodes, especially with respect to their switching characteristics because they have negligible reverse recovery losses. SiC MOSFETs and JEFTs with reasonably low on-resistance were reported [24], which can potentially be used as the active power switches replacing Si-IGBTs. It is a great challenge to produce a high-quality interface between silicon carbide and a suitable gate dielectric material for SiC MOSFET structures. In addition to a large density of charge in the oxide and at the interface that causes threshold voltage shift, the inversion channel mobility was found to be very low when compared with silicon. To compound the problem, the conventional silicon power MOSFET structure could not provide the full benefits of the high breakdown field strength of SiC material because of reliability and rupture problems associated with the gate oxide. An SiC JFET does not encounter these problems due to the absence of the gate dielectric layer but has the disadvantage of requiring a negative gate voltage to block voltages. This disadvantage can be overcome by combining a normally on SiC high-voltage JEFT with a low-voltage silicon power MOSFET to create a hybrid switch with the desired features for inverter applications [24]. Another approach to overcome the gate dielectric issue is to shield the channel region by formation of either a P-type region or by creating a high resistivity conduction barrier region under the channel [24]. In particular, the shielded accumulationmode MOSFET structures (named ACCUFET) demonstrated promising characteristics for the development of monolithic power switches from silicon carbide [25], [26]. The unique material property of the WBG semiconductor may enable novel circuit topology and control schemes to improve the efficiency of the inverter. One example is to replace the freewheeling diode with a BWBG

synchronous rectifier.[ The diode voltage drop due to the potential barrier (0.6–1.1 V), such as Schottky barrier or PN junction built-in potential, will remain to be the major semiconductor loss in the future while the switching device conduction characteristics will be improved. The WGB switches will remove the potential barrier losses by utilizing the reverse conducting characteristics of the devices. During the current commutation phase of the inverter circuit, the gate of the WBG switch can be turned on so that the current flows through the switching device in reverse direction instead of the loss generating antiparallel freewheeling diode. Although the synchronous rectifier technique has been widely used in low voltage dc–dc converters, high voltage silicon MOSFETs and IGBTs are not suitable to this technique because of the parasitic body diode action of silicon MOSFETs and the reverse current blocking characteristics of IGBTs. In WBG unipolar devices, such as SiC-MOSFETs, the parasitic body diode action is prevented due to the material property so that the conduction losses during current commutation phase will be significantly reduced. The impact of SiC devices in hybrid electric vehicles was investigated [27]–[29]. A 55-kW SiC–Si hybrid HEV traction inverter with Si IGBTs and SiC Schottky diodes was built, characterized, and compared with a similar all-Si inverter [29]. When SiC Schottky diodes are used in HEV powertrain inverters replacing Si PN diodes, the average losses of the inverter can be reduced by up to 33.6% [30]. A 7.5-kW all-SiC inverter with SiC JFETs and Schottky diodes was also built, characterized, and compared with a similar all-silicon inverter [30]. It is shown that the result is promising but more research is required on SiC JFETs to improve their performance and reliability. There are still many technical and economical barriers that prevent SiC technology from large-scale commercialization in the near future. Many technical barriers, including the high defect density and cost of SiC wafers, exist and prevent SiC from wide-spread commercialization. Currently, the cost of SiC devices ranges from five to ten times that of silicon devices with the same voltage and current ratings. It is projected that the cost ratio between SiC and silicon will decrease to three in three to five years and eventually to one in eight to 12 years. Unlike military or space applications, automotive applications are extremely sensitive to component costs. One critical question is how to fully exploit these benefits to achieve system-level improvement in the weight, size, efficiency, performance, and even possibly the overall cost of the power electronic subsystem. In other words, the dramatic increase in component costs due to the introduction of SiC devices must be fully justified. We must rethink the ways power electronics are designed and constructed in electric vehicles, including but not limited to, converter topology, switching frequency, maximum operating temperature, thermal management, system partitioning, and package/ assembly techniques. Vol. 95, No. 4, April 2007 | Proceedings of the IEEE

785

Shen and Omura: Power Semiconductor Devices for Hybrid, Electric, and Fuel Cell Vehicles

VI . ADVANCED POWER MODULE TECHNOLOGY Multichip hybrid power modules are predominantly used in hybrid electric vehicles as well as industrial applications. The hybrid modules distribute signal and power, dissipate heat, protect the devices enclosed, and serve as the basic power electronics building block (PEBB). Power module design must also relieve the inherent mechanical stresses after bonding large silicon IGBT chips having low coefficient of thermal expansion (CTE) to other materials having higher CTE. Automotive modules must function down to 40  C where CTE mismatch stresses become excessive. The state-of-the-art power module technology, initially developed in the mid 1980s, mainly relies on the use of aluminum wirebonds, direct-bond-copper (DBC) ceramic substrates, and copper base plates, as shown in Fig. 6. The thin aluminum wirebonds suffer from high parasitic impedance, fatigue-induced lift-off failures, and inability to remove heat. The DBC ceramic substrate (Al2 O3 or more expensive AlN) provides electrical isolation but inadvertently increases the package thermal resistance. The thick Cu base plate serves as a heat spreader but considerably increases the weight, size, and thermal resistance of the power module. The temperature of the coolant, used to remove heat from the power electronics in HEV applications, is being pushed higher, toward that of the engine coolant. However, at higher coolant temperatures, the usable current, needed by such power semiconductor devices as IGBTs and diodes, becomes severely limited. The problem is the

Fig. 6. Conventional IGBT power module structure.

786

Proceedings of the IEEE | Vol. 95, No. 4, April 2007

inefficient heat dissipation capability of existing power module packaging materials, a problem that becomes even more exaggerated as the die size of IGBT devices decreases due to improvements in device structure. The issues of a conventional power module are mainly due to the complexity of its material system, which is comprised of multiple semiconductor chips, one or more ceramic substrates with patterned metal interconnect films, a metal baseplate, Al or Au wirebonds connecting the semiconductor chips to the substrate metal layer, soldering joints of semiconductor-to-substrate and substrate-to-baseplate, epoxy sidewall module case, external metal posts or feedthroughs, and silicon-gel potting material. Such a module construction method unnecessarily increases the module’s weight, size, and junction-to-case thermal resistance. More critically, the complex material system creates many joints and interfaces between dissimilar materials, which tends to cause reliability concerns at high temperatures or under thermal cycling conditions [31]–[33]. For example, it is well known that aluminum wirebonds tend to fail by creep deformation at high temperatures, and Sn/Pb solder joints suffer from voids or delamination under thermal cycling in the conventional power modules [34]–[36]. While the reliability issue can be mitigated to a certain extent through the selection of new materials and processes (often with the penalty of a sharp increase in cost), it would be much more advantageous to begin with a simplified material system. Furthermore, in the conventional hybrid power modules, heat is solely dissipated from the semiconductor

Shen and Omura: Power Semiconductor Devices for Hybrid, Electric, and Fuel Cell Vehicles

Fig. 7. Evolution of power packaging technology.

chips downward to the ceramic substrate, metal baseplate, and eventually to the external thermal management components such as the heat sinks or liquid-cooling cold plates. Note that the top side of the power module does not contribute at all to removing heat from the semiconductor chips. More disruptive technological innovation in both module architecture and material selection is needed to provide smaller, lighter, cheaper, and more reliable modules for high temperature power electronics in both military and commercial applications. Several novel packaging concepts have been proposed to overcome the aforementioned issues, including but not limited to, removal of wirebonds and double-side cooling of the power chips, integration of cooling plates into package, and increase of maximum junction operating temperature. Double-side cooling techniques were employed in pressure contact packaged IGBTs for utility, traction, and high power industrial applications [37]. Its feasibility for IGBT module was also investigated in several recent

publications. IGBT emitter contacts can be made through a metal block which improves heat dissipation characteristics of the power module [38], [39]. The reliability issue for the metal contact is also discussed [40], [41]. Techniques of integrating cooling plates into IGBT modules was also investigated and showed that thermal resistance can be reduced by more than 60% [42]. Extremely high power dissipation flux over 400 W/cm2 was demonstrated with an IGBT chip directly sandwiched by cooling plates [43]. As a future possibility, chip direct cooling with liquid splaying or micro channel on/in the semiconductor shows power flux in the range of 200 W/cm2 or higher [44]–[47]. Spray cooling with fluorinert liquid has already showed more than 100 W/cm2 of heat dissipation flux for single side cooling and 200 W/cm2 for double-side cooling. This technology shows a great potential for further improvement, such as nozzle shape, liquid flow management, and dielectric liquid property. Fig. 7 summarizes the chip cooling technology on a trend graph of estimated chip power loss as a function of chip current density. Vol. 95, No. 4, April 2007 | Proceedings of the IEEE

787

Shen and Omura: Power Semiconductor Devices for Hybrid, Electric, and Fuel Cell Vehicles

VI I. SUMMARY Power semiconductor devices are a key factor in determining the efficiency, performance, and cost of hybrid, electric, and fuel cell vehicle systems. Even though the semiconductor industry has made impressive progress in advancing power device technology in the past two decades, there is still plenty of room for further improvement in both device and package performance. On one hand, SiC and other wide bandgap semiconductors have shown great promise in replacing silicon IGBTs and/or diodes in powertrain inverters to offer a higher efficiency. On the other hand, the performance of silicon devices has yet to reach their full potential. The theoretical current

REFERENCES [1] A. Kawahashi, BA new-generation hybrid electric vehicle and its supporting power semiconductor devices,[ in Proc. 2004 Int. Symp. Power Semiconductor Devices and ICs, pp. 23–29. [2] D. Himilton, 2001 Annual Progress Report: Vehicle Power Electronics and Electric Machines, U.S. Dep. Energy, Dec. 2001. [3] V. R. Stefanovic, BTrends in AC drive applications,[ in Proc. Int. Power Electronics Conf. CD-ROM, 2005. [CD-ROM]. [4] K. Hamada, BPower electronics technologies support new generation automobiles,[ in Proc. Int. Workshop Power Electronics New Wave, National Institute of Advanced Industrial Science and Technology (AIST), 2005, pp. 73–81. [5] J. M. Miller, BHybrid vehicle propulsion: Exploring the electronic CVT,[ in Proc. Annu. Applied Power Electronics Conf. Expo. APEC 2005 Professional Education Seminars Workbook, 2005, vol. 3, p. S18. [6] ENNA/NEDO, Ultra Low Loss Power Semiconductor Development Survey on Practical Use of Next Generation Power Semiconductor Device, 2003, (in Japanese). [7] I. Omura and G. Majumdar, BCurrent status and future prospect of semiconductor power devices,[ in Proc. Int. Workshop Power Electronics New Wave, National Institute of Advanced Industrial Science and Technology (AIST), 2005, pp. 53–70. [8] A. Nakagawa, H. Ohashi, M. Kurata, Y. Yamaguchi, and K. Watanabe, BNon latch-up 1200 V 75 A bipolar-mode MOSFET with large ASO,[ in IEDM Tech. Dig., 1984, pp. 860–861. [9] A. Nakagawa and H. Ohashi, B600- and 1200-V bipolar-mode MOSFET’s with high current capability,[ IEEE Electron Device Lett. vol. EDL-6, no. 7, pp. 378–380, Jul. 1985. [10] G. Miller et al., BA new concept for a non punch through IGBT with NOSFET like switching characteristics,[ in Power Electronics Specialists Conf., 1989. PESC ’89 Rec., 1989, vol. 1, pp. 21–25. [11] G. Majundar, H. Sugimoto, M. Kimata, T. Iida, H. Iwamoto, T. Nakajima, and H. Matsui, BSuper mini type integrated inverter using intelligent power and control devices,[ in Proc. 2004 Int. Symp. Power Semiconductor Devices and ICs, 1990, pp. 144. [12] M. Kitagawa, I. Omura, S. Hasegawa, T. Inoue, and A. Nakagawa, BA 4500 V injection enhanced insulated gate bipolar transistor (IEGT) operating in a mode similar

788

[13]

[14]

[15]

[16]

[17]

[18]

[19]

[20]

[21]

[22]

handling capability of the silicon BnanoIGBT[ exceeds not only that of today’s IGBT products but also that of the best performing SiC and GaN wide bandgap semiconductor devices reported recently. As nanofabrication technology becomes more available and affordable, we can expect to see the commercial success of the silicon nanoIGBT concept. It would be interesting to observe how resilient the ever-evolving silicon technology is against the competition from the wide bandgap semiconductors in HEV applications. Regardless of the outcome of this competition, the automotive industry and ultimately the consumers of future hybrid, electric, and fuel cell vehicles will be sure winners. h

to a thyristor,[ in IEDM Tech. Dig., 1993, pp. 679–682. I. Omura, T. Ogura, T. Sugiyama, and H. Ohashi, BCarrier injection enhancement effect of high voltage MOS devices-device physics and design concept,[ in Proc. Int. Conf. Power Semiconductor Devices and IC’s, 1997, pp. 217–220. M. Harada, T. Minato, H. Takahashi, H. Nishihara, K. Inoue, and I. Takata, B600 V trench IGBT in comparison with planar IGBT-an evaluation of the limit of IGBT performance,[ in Proc. Int. Conf. Power Semiconductor Devices and IC’s, 1994, pp. 411–416. H. Takahashi, H. Haruguchi, H. Hagino, and T. Yamada, BCarrier stored trench-gate bipolar trensistorVA novel power device for high voltage application,[ in Proc. Int. Conf. Power Semiconductor Devices and IC’s, 1996, pp. 349–352. D. Burns, I. Deram, J. Mello, J. Morgan, I. Wan, and F. Robb, BNPT-IGBTV Optimizing for manufacturability,[ in Proc. Int. Conf. Power Semiconductor Devices and IC’s, 1996, pp. 331–334. T. Laska, M. Matschittsch, and W. Scholz, BUltrathin-wafer technology for a new 600V-NPT-IGBT,[ in Proc. Int. Symp. Power Semiconductor Devices and ICs, 1997, pp. 361–364. T. Matsudai, K. Kinoshita, and A. Nakagawa, BNew 600 V trench gate punch-through IGBT concept with very thin wafer and low efficiency p-emitter, having an on-state voltage drop lower than diodes,[ in Proc. IPEC, 2000, pp. 292–296. T. Laska, M. Munzer, F. Pfirsch, C. Shaeffer, and T. Schmidt, BThe field stop IGBT (FS IGBT)VA new power device concept with a great improvement potential,[ in Proc. Int. Symp. Power Semiconductor Devices and ICs, 2000, pp. 355–358. K. Nakamura, S. Kusunoki, H. Nakamura, Y. Ishimura, Y. Tomomatsu, and T. Minato, BAdvanced wide cell pitch CSTBT having Light Punch-Through (LPT) structures,[ in Proc. Int. Symp. Power Semiconductor Devices and ICs, 2002, pp. 277–280. A. Nakagawa, BTheoretical investigation of silicon limit characteristics of IGBT,[ in Proc. Int. Conf. Power Semiconductor Devices and IC’s, 2006, pp. 5–8. M. Baus, B. N. Szafranek, S. Chmielus, M. C. Lemme, B. Hadam, B. Spangenberg, R. Sittig, and H. Kurz, BFabrication of monolithic bidirectional switch (MBS) devices with MOS-controlled emitter

Proceedings of the IEEE | Vol. 95, No. 4, April 2007

[23]

[24] [25]

[26]

[27]

[28]

[29]

[30]

[31]

[32]

[33]

[34]

structure,[ in Proc. Int. Conf. Power Semiconductor Devices and IC’s, 2006, pp. 181–184. V. K. Khanna, IGBT Theory and Design. New York: Wiley-Interscience, 2003, pp. 60–62. B. J. Baliga, Silicon Carbide Power Devices. Singapore: World Scientific, 2005. P. M. Shenoy and B. J. Baliga, BThe planar 6H-SiC ACCUFET,[ IEEE Electron Device Lett., vol. 18, pp. 589–591, 1997. S. Kaneko et al., B4H-SiC ACCUFET with a two-layer stacked gate oxide,[ in Silicon Carbide and Related Materials-2001, Material Science Forum, 2002, vol. 389–393, pp. 1073–1076. L. M. Tolbert, B. Ozpineci, M. Chinthavali, S. K. Islam, and F. Z. Peng, BImpact of SiC Power electronic devices for hybrid electric vehicles,[ SAE Trans. J. Passenger Cars-Electronic Electrical Systems, pp. 771–765, 2003. B. Ozpineci, M. Chinthavali, and L. M. Tolbert, BA 55 kW three-phase automotive traction inverter with SiC Schottky diodes,[ in IEEE Vehicle Propulsion and Power Conf. (VPPC), Chicago, IL, Sep. 7–9, 2005. M. Chinthavali, L. M. Tolbert, and B. Ozpineci, BHigh temperature and high frequency performance evaluation of 4H-SiC VJFETs and Schottky diodes,[ in Proc. IEEE Applied Power Electronics Conf. (APEC), Austin, TX, Mar. 8–10, 2005. S. Rogers, 2005 Annual Progress Report for the Advanced Vehicle Power Electronics and Electric Machines, U.S. Dep. Energy, Nov. 2005. N. Seliger, E. Wolfgang, G. Lefranc, H. Berg, and T. Licht, BReliable power electronics for automotive applications,[ Microelectronics Reliability, vol. 42, pp. 1597–1604, 2002. W. Kanert, H. Dettner, B. Plikat, and N. Seliger, BReliability aspects of semiconductor devices in high temperature applications,[ Microelectronics Reliability, vol. 43, pp. 1839–1846, 2003. R. Zehringer, A. Stuck, and T. Lang, BMaterial requirements for high voltage, high power IGBT devices,[ Solid-State Electron., vol. 42, pp. 2139–2151, 1998. K. Sommer, J. Gottert, G. Lefranc, and R. Spanke, BMultichip high power IGBT-modules for tranction and industrial application,[ in Proc. Eur. Conf. Power Electronics and Applications (EPE’97), 1997, vol. 1, pp. 1.112–1.116.

Shen and Omura: Power Semiconductor Devices for Hybrid, Electric, and Fuel Cell Vehicles

[35] M. Ciappa, BSome reliability aspects of IGBT modules for high-power applications,[ Ph.D. dissertation, ser. Microelectronics, vol. 111, Hartung-Gorre Verlag, Konstanz, ISBN 3-89649-657-32001. [36] M. Ciappa, BLifetime modeling of thermomechanics-related failure mechanisms in high power IGBT modules for traction applications,[ in Proc. Int. Symp. Power Semiconductor Devices and ICs, 2003, pp. 295–298. [37] H. Matsuda, M. Hiyoshi, and N. Kawamura, BPressure contact assembly technology of high power devices,[ in Proc. Int. Symp. Power Semiconductor Devices and ICs, 1997, pp. 17–24. [38] M. Otsuki, H. Kanemaru, Y. Ikeda, K. Ueno, M. Kirisawa, Y. Onozawa, and Y. Seki, BAdvanced thin wafer IGBTs with new thermal management solusion,[ in Proc. International Symp. Power Semiconductor Devices and ICs, 2003. [39] A. Narazaki, T. Shirasawa, T. Takayama, S. Sudo, N. Asano, N. Asano, K. Ogura,

[40]

[41]

[42]

[43]

[44]

H. Takahashi, and T. Minato, BDirect beam lead bonding for trench MOSFET & CSTBT,[ in Proc. Int. Symp. Power Semiconductor Devices and ICs, 2005, pp. 75–78. M. Usui, M. Ishiko, K. Hotta, S. Kuwano, and M. Hashimoto, BEffects of uni-axial mechanical stress on IGBT characteristics,[ Microelectronics Reliability, vol. 45, pp. 1682– 1687, 2005. H. Tanaka, K. Hotta, S. Kuwano, M. Usui, and M. Ishiko, BMechanical stress dependence of power device electrical characteristics,[ in Proc. Int. Symp. Power Semiconductor Devices and ICs, 2006, pp. 1–4. J. S. Harder, K. Exel, and A. Meyer, BDirect liquid cooling of power electronics devices,[ in Proc. 4th Int. Conf. Integrated Power Systems, 2006, pp. 347–351. T. Steiner and R. Sittig, BIGBT module setup with integrated micro-heat sinks,[ in Proc. Int. Symp. Power Semiconductor Devices and ICs, 2000. K. Vanam, J. Junghans, F. Barlow, R. P. Selvam, J. C. Balda, and A. Elshabini,

BA novel packaging methodology for spray cooling of power semiconductor devices using dielectric liquids,[ in Annu. Applied Power Electronics Conf. Expo. APEC, 2005. [45] M. J. Ellsworth and R. E. Simons. (2005, Aug.). ‘‘High powered chip coolingVAir and beyond,[ in Electronics Cooling. [Online]. Available: http://www. electronics-cooling.com/. [46] R. S. Prasher, J. Y. Chang, I. Sauciuc, S. Narasimhan, D. Chau, G. Chrysler, A. Myers, S. Prstic, and C. Hu. (2005). Nano and micro technology-based next-generation package-level cooling solutions. Intel Technol. J. [Online]. 9(4), pp. 285–296. Available: http://www. intel.com/. [47] G. Mitic, W. Kiffe, G. Lefranc, and S. Ramminger, BComparison of spray cooling with direct liquid base-plate flow convection of IGBT power modules,[ presented at the EPE’02, Paper SSJAF-04.

ABOUT THE AUTHORS Z. John Shen (Senior Member, IEEE) received the B.S.E.E. from Tsinghua University, China, in 1987, and the M.S. and Ph.D. degrees from Rensselaer Polytechnic Institute, Troy, NY, in 1991 and 1994, respectively, all in electrical engineering. He is an Associate Professor of Electrical Engineering at the University of Central Florida, Orlando (UCF). Between 1994 and 1999, he held a variety of positions including Senior Principal Staff Scientist with Motorola Semiconductor Products Sector, Phoenix, AZ. He was with the University of Michigan, Dearborn, as an Assistant and Associate Professor in the Department of Electrical and Computer Engineering between 1999 and 2003. His research interests include power electronics, power semiconductor devices and ICs, power management for computers and telecom equipment, automotive electronics, renewable and alternative energy systems, and electronics manufacturing. He has published over 80 journal and conference articles, and holds nine issued and several pending U.S. patents in these areas. Dr. Shen is a recipient of the 2003 NSF CAREER Award, the 2003 IEEE Best Automotive Electronics Paper Award from IEEE Society of Vehicular Technology, and the 1996 Motorola Science and Technology Award. He currently serves as the Automotive Power Electronics Technical Committee Chair and at-large AdCom member of the IEEE Power Electronics Society. He also serves as an Associate Editor of IEEE TRANSACTIONS IN POWER ELECTRONICS.

Ichiro Omura (Member, IEEE) received the M.S. degree in mathematics from Osaka University, Osaka, Japan, in 1987 and the Ph.D. degree in electrical engineering from the Swiss Federal ¨ rich, Switzerland, Institute of Technology (ETH), Zu in 2001. From 1987 to 1999, he was with the Research and Development Center, Toshiba Corporation, Kawasaki, Japan. From 1996 to 1997, he was a Visiting Researcher at ETH. Since 1999, he has been with the Advanced Power Semiconductor Device Development Group, Semiconductor Company, Toshiba Corporation, where he has been engaged in research and development of power semiconductor devices. Dr. Omura is a member of the Institute of Electrical Engineers of Japan.

Vol. 95, No. 4, April 2007 | Proceedings of the IEEE

789