The Creation of Silicon Carbide - Revolutionary Semiconductor

The Creation of Silicon Carbide - Revolutionary Semiconductor The History of Silicon Carbide Silicon carbide (SiC) is a rather young base material in the semiconductor industry compared to silicon or gallium arsenide, but its origins date back to the end of the 19th century. In 1891, Edward Acheson developed a method for producing crystalline SiC as an abrasive material, a method still in use today. The first mineralogical studies of the secondhardest mineral occurring in nature carried out by the French chemist H. Moissan also date back to this time. SiC is sometimes still called Moissanite today. By Dr. Robert Eckstein, CEO SiCystal AG and Aly Mashaly, Manager Power Systems Department; Rohm Semiconductor GmbH When the semiconductor age began in the middle of the 20th century with the introduction of the bipolar transistor, the enormous potential of silicon carbide for specific applications was already understood. According to the discoveries of E. Johnson, the product of breakdown field strength and saturation drift velocity may be used as an indication of a material s suitability as a base material for transistors in RF and power applications. In this comparison, SiC turns out to be much better than both silicon and gallium arsenide. In practical applications however, the innovative semiconductor material was unable to prove its superiority immediately. No process existed at that time able to produce sufficiently large silicon carbide substrates with the low defect rates required to manufacture meaningful test structures. Following on from the Acheson process and the so-called Lely process as an intermediate step, it was the modified Lely process that provided the basis for the modern production of SiC substrates at the end of the 1970 s. When better and larger silicon carbide substrates became available thanks to the evolution of this basic process, SiC provided a completely convincing performance. The source material, silicon carbide, acts as raw material in the growth process and continuously feeds the growing crystal. The source material normally is in powder form, and it is produced in a preceding process. This process is a strictly controlled high-temperature process combining the base materials which are highly pure silicon and carbon. Precise control ensures that base material is used which is characterized by a specific, internally specified purity, stoichiometry, grain size, grain distribution and other internally defined specified parameters. Prerequisites for Producing Silicon Carbide Crystals Before the production (which is called growth by experts) of a silicon carbide crystal can begin, the essential components must be in place. These include the crystal growth equipment itself, the growth process, various media, the source material and the seed crystal. Figure 1: Significance of the Seed Crystal for Growing Si and SiC Crystals www.bodospower.com January - March 2017 Bodo s Power Systems 1

The seed crystal is a mono-crystalline, round silicon carbide slice that initiates (and largely controls) the growth of a large mono-crystal, hence the term seed. Like the seed of a plant, the seed crystal imparts significant genetic information to the growing crystal. The structural perfection of the seed crystal determines much of the crystal s optimum absence of defects. The multi-stage selection process used in our company continuously ensures that seed crystals will only be made of the best in-house crystals featuring optimum material quality. This constant internal selection acts like a distillery process continuously improving the quality of the seed and the final product. During crystal growth, a doping material (normally gaseous nitrogen) is intentionally added in the vapor phase. Incorporated into the crystal in a low, well-defined quantity, the dopant electrically acts as a donor ensuring the intended conductivity of the crystal. The Significance of the Seed Crystal as Confinement to the Growth of Silicon Crystals The crystal growth equipment, the growth process, the media, the source material and the seed crystal play a significant role in the processes used for producing other semiconductors as well (including silicon). However, silicon carbide is different! As explained in more detail in the section below, SiC crystals are grown by a special vapor-phase process. For this reason, the diameter of the seed crystal should have at least the target diameter of the crystal to be grown. It is not possible to grow a large crystal of sufficient quality on a seed rod with a relatively small diameter, as would be the case with silicon (see Figure 1). Any defects in the seed would be inherited by the growing crystal due to the genetics described above, mandating the use of a flawless, large-diameter seed crystal for silicon carbide. Therefore, the seed crystals are among the factors supporting the secure operation of the enterprise. Not commercially available anywhere, this material acts as a kind of copy protection. Securely protected by internal processes, it is always available in sufficient quantity at SiCrystal. The Growth of the Silicon Carbide Crystal When all the required conditions are met as described above, the growth process itself can begin. The sublimation growth method is usually used for SiC. The setup then consists of a graphite crucible, which is surrounded by carbon-based insulation material. This helps to save energy and is inserted into the induction-heated growth equipment (see Figure 2). The crystal growth process then proceeds in the following manner: The induction coil of the growth equipment induces circular currents in the graphite growth crucible, heating it up to well over 2273 K. At these elevated temperatures, the source material in the lower zone of the crucible evaporates, the silicon carbide powder sublimes, turning into various gaseous species that consist of silicon and/or carbon which fills up the free space of the crucible. The seed crystal attached to the top of the crucible does not evaporate because temperatures are lower there, although they still are above 2273 K. Adhering to the laws of thermodynamics, the gaseous species formed by the source material at temperatures far above 2273 K have the tendency to deposit from the vapor phase when they are transported close to the colder seed crystal. Crystallizing on the surface of the seed, they form additional crystal volume and let the crystal grow. However, the crystal does not grow in an uncontrolled manner because the genetics of the seed predicts the growth with a crystallographic relationship to the seed. Figure 2: SiC Crystal Growth Process The growth process therefore relies on the constant transfer of material from the source to the growing crystal. The growth process, which requires the continuous transport of species and dopant through the vapor phase, lasts a couple of days. When the powder has almost completely evaporated and a crystal of several centimeters in length has grown, the process is terminated in a controlled manner. Several other aspects must be considered which enable the process in the way described above. As already mentioned, the crucible and its insulation are made of carbon. Carbon lends itself to this purpose because of its high temperature stability, its availability in semiconductor purity and the fact that it does not represent an elementary contamination for SiC. However, all carbon parts would immediately burn at 2273 K under normal atmospheric conditions. Therefore, it is necessary to run the process in an inert atmosphere (Argon is often used). To contain all gases including the Argon, the crucible and its insulation are surrounded by two coaxial glass tubes whose intermediate space is cooled with water. Manufacturing Silicon Carbide Substrates: The Wafer-Production Process Chain At the end of the crystal growth process, a cylindrical single crystal of several centimeters in length, whose diameter is larger than the target diameter, can be taken from the crucible. The subsequent process steps are almost identical to other established semiconductors, including silicon. The main steps include: Crystal orientation using x-ray analysis Cylindrical grinding of the crystal to the desired diameter Grinding of flats (markings on the crystal surface) Sawing wafers from the crystal Application of individual laser markings for wafer identification Rounding off the edges of the wafers Surface treatment with methods including Grinding Lapping Polishing by mechanical abrasion Polishing by chemical/mechanical abrasion Cleaning the wafers using dry or wet chemical methods 2 Bodo s Power Systems January - March 2017 www.bodospower.com

As a matter of course, these process steps are accompanied by checking many internally or externally specified parameters using special optical, mechanical, chemical and electrical methods. The large amount of data that is acquired and archived ensures full traceability of the final product back even to the raw material batch used in the initial process stage. The process described above is long established as the standard in manufacturing semiconductor substrates. The process stages, processing facilities and inspection methods are basically known from silicon processing. Nonetheless, processes and/or machines sometimes had to be modified to accommodate the specific properties of silicon carbide. Silicon carbide is a very hard material. On Mohs hardness scale, where diamonds represent the upper limit with a hardness of 10 and all minerals are considered very hard with hardness values greater than 6, SiC is among the top scorers with a hardness of 9. Only a few materials are available for abrasive machining of silicon carbide. Tools and sometimes even process equipment has to be adapted to reliably provide semiconductor customers the required product precision. Growth Process and Wafer Production Challenges and Solutions Not surprisingly, ensuring highest quality based on ever tightening specifications is among the key challenges in the production of silicon carbide crystals and substrates. In this context, quality means the quality of the wafer surface: Absence of scratches Extremely low, nanometer-scale surface roughness Surface purity in a physical sense (particles, films) and in an electrical sense (electrically active impurities) Comparisons of old and current substrates using stress birefringence method reveal the enormous improvements of material quality in the last 15 years or so. Stress birefringence is an optical method representing deviations from the ideal state as white areas, while undisturbed regions are displayed as black areas in those pictures shown below. Strong contrasts are visible on the entire surface of the wafer from the year 2000, which is shown on the left (see Figure 3), indicating defectrelated stress. Responsible defects include dislocations and related structural defects, proving that the growth process was not sufficiently optimized in the year 2000. The current state of the art shown ( Figure 3.) on the right is quite different. Although the sensitivity of the optical method has improved a lot in the meantime, there are almost no contrasts between dark and bright areas. This clearly indicates that the growth process was optimized significantly, that the defect density within the crystal has been reduced by several orders of magnitude and that current wafers are production-worthy without any limitation. But how can these improvements of material quality be achieved? In this case, these quality improvements were obviously achieved by an optimization of the crystal-growth process. But how can one optimize a high-temperature process operating at temperatures far above 2273 Kelvin Placing sensors in the crucible area is hardly conceivable at these elevated temperatures. Furthermore, any additional observation hole in the insulation for optical temperature measurements from the outside would unacceptably influence the temperature distribution within the crucible. Any macroscopic opening in the growth crucible would also inevitably lead to the escape of process gases, which would corrupt the growth process. In addition, it also means geometrical quality: Tight diameter and thickness tolerances Minimum bow and warp of substrates Exactly specified edge shape Also, to a high degree, it means internal or better intrinsic quality: Homogeneity of the doping and thus the conductivity Stresses in the material that are as low as possible Low density of dislocations or related defects Today s State of the Art of SiC Quality Although quality naturally is a fast moving target, it is justified to say that SiC fully meets the customer s requirements for an established semiconductor material today. However, much work was required in the past to reach and maintain this level. Figure 4: Temperature distribution in the crystal growth crucible The problem can be solved by numerically simulating the growth of the crystal (see Figure 4). Using special software and considering all relevant material and process parameters, crystal growth experts now can look into the crucible. Figure 3: Diameter increase and quality improvement Experienced crystal growers adequately recognize from the temperature distribution (the so-called thermal field) shown above how a crystal would grow under modified growth conditions. A multitude of numerical experiments will limit the number of necessary real experiments. The combinations of simulation with experimental confirma- www.bodospower.com January - March 2017 Bodo s Power Systems 3

tion have enabled crystal growers with many years of experience to achieve the radical improvement of material quality described above. Outlook The above mentioned comparison between a current product and the wafer from the year 2000 also reveals the diameter increase from 35mm to 150mm, which was achieved during the intervening years. As can be seen, today s wafers feature a significant quality improvement on a wafer surface which has increased by a factor of 18. This trend towards increased wafer size accompanied by continuously improving quality will continue as the most important goal in the future. Predominantly as larger wafer surfaces can be processed more economically, thereby lowering production costs. crystallizing on a SiC seed crystal. This creates a single crystal which is subsequently processed into substrates using modified process stages long established in the semiconductor industry. Continuously increasing requirements regarding the quality and diameter of the substrates were and are considered by process adaptations based on computer simulation. Thanks to their high quality level, today s substrates with diameters of up to 150mm can be used for component production without limitation. Optimally positioned for the present and the future, the ROHM subsidiary SiCrystal is prepared for the journey towards larger diameters and further improved quality. About SiCrystal AG SiCrystal AG is among the world s leading manufacturers of silicon carbide semiconductor substrates. Founded in 1996 and headquartered in Nuremberg (Germany), the wholly owned subsidiary of ROHM CO. LTD delivers its products to customers all over the world. The company owns all processing stages from the production of the original material to packaging the epi-ready substrate in an in-house clean-room environment. In addition to the results of its comprehensive, proprietary analysis methods, SiCrystal AG has direct access to the results of component tests within the ROHM Group. This is an invaluable benefit for advancing the material optimization based on component or even module performance. www.rohm.com /eu Figure 5: Evolution of SiC wafer diameters While the transition from 100mm to 150mm has just begun for today s customers, there is much interest in the possible introduction of the 200mm substrate as the next stage of evolution. Although it will take some time for this next step to become commercially viable, SiCrystal expects to be able to deliver the first samples within two or three years. Consequently, preliminary work on this has already started. Dr. Robert Eckstein, CEO SiCystal AG Summary Although the potential of silicon carbide for power and RF electronics applications has been recognized quite early, a suitable manufacturing process for this material was only developed by the end of the 1970 s. In the industrial process used until today, the original SiC material sublimes before being transported through the vapor phase and Aly Mashaly, Manager Power Systems Department Rohm Semiconductor GmbH SMALLER STRONGER FASTER www.rohm.com/eu 4 Bodo s Power Systems January - March 2017 www.bodospower.com

The Creation of SiC - Cell Structures and Production Process The first part of this article series described the creation of Silicon Carbide (SiC) substrate wafers, starting with the production of the raw material (SiC powder) to the so-called epi-ready SiC substrate wafers. By Aly Mashaly, Manager Power Systems Department, Rohm Semiconductor GmbH And Mineo Miura, SiC Power Device Engineer, Rohm Co., Ltd. The second part deals with the potential structures of SiC devices, focusing on different structures including SiC Schottky Barrier Diodes (SBDs), planar SiC MOSFETs and double trench MOSFETs, showing that cell structures significantly variate the physical properties and the performance of the final product. Finally, production testing procedures followed by Rohm for quality assurance purposes are discussed. Introduction AC/DC, DC/AC, DC/DC and AC/AC converters are the most popular power electronic systems. The efficiency of conventional power electronics technologies usually varies between 85 and 95 percent. In other words, approximately 10% of electrical energy is dissipated in form of heat at each power conversion stage. Generally speaking, the performance of power semiconductors is considered to be the main limiting factor for the efficiency of power electronics. Therefore, developing high-voltage and low-loss power semiconductors is a prerequisite for building the power grids of the future. According to laws of semiconductor physics, the specific on resistance (RonA) increases dramatically with the breakdown voltage. Due to the above mentioned properties of SiC, the RonA value of SiC will be 100 times lower at high voltages compared to Si. Because of its low RonA characteristics at high voltage, there is no need for using minority carrier device structure normally used in silicon high voltage devices like IGBTs and FRDs. In SiC power devices, majority of carrier devices like MOSFETs and SBDs are used for 600 to 3.3kV voltage range. Due to the absence of minority carriers in current conduction the switching speed of SiC is dramatically improved, which lead to a dramatic reduction in switching losses. The mentioned features make SiC a very promising material for high-voltage applications, where thermal management is particularly important. Compared to silicon (Si) semiconductors, the electrical field strength of SiC is almost ten times higher (2.8MV/cm vs. 0.3MV/cm). The increased field strength of SiC material enables the deposition of a thinner layer structure, which is known as epitaxial layers on the SiC substrate. Its thickness is about one tenth of that of Si epitaxial layers. At the same breakdown voltage, the doping concentration of SiC can be two orders of magnitude higher than of its Si equivalent. This reduces the device s specific on resistance (RonA), resulting in a significant reduction of its conduction losses (Figure 1). Figure 2: Specific on resistance (RonA) vs. breakdown voltage Figure 1: The thinner layer stack enabled by the increased breakdown field strength of SiC leads to lower conduction losses Figure 3: Many physical properties make SiC an attractive proposition for power electronics applications www.bodospower.com January - March 2017 Bodo s Power Systems 5

Challenges facing power electronic systems have increased significantly in recent years. Requirements such as weight and efficiency play a predominant role. Furthermore, total system costs and efforts must be low during production phase, without degrading the quality and robustness of the end product. Thanks to its physical properties, SiC has an enormous potential to meet the requirements of these challenges and the associated market trends. In power electronic systems, thermal design plays an important role in achieving systems featuring high power density and compact size. SiC is perfectly suited for these applications, as it offers a three times better thermal conductivity than Si semiconductors. In addition, SiC supports higher operating temperatures (even over 250 C is in principle possible) compared to Si semiconductors because of its wider band gap (three times of Si). SiC Diodes Compared to Si diodes, SiC SBDs are much more attractive for power electronic applications especially at voltages of 600V and beyond. SiC SBDs feature much better efficiency due to their lower switching losses and the elimination of the so-called reverse recovery current during turn-off (see Figure 5). The EMI performance of the entire system is improved because EMI emissions are reduced accordingly. The Production of a SiC Device The SiC substrate wafer was described in detail in part 1 of this article series. These substrate wafers act as the base material for the subsequent production of SiC devices. Specific structures consisting of epitaxial layers, doping processes and metallization finally produce a SiC device, which can be a SiC diode, a SiC MOSFET or even a SiC IGBT depending on the specific structures. During the development of SiC devices, a highly precise epitaxial growth process is an essential prerequisite for producing drift layers with the desired thickness and optimum doping concentration. Using process optimization and adequate cleaning and conditioning of polished surface of substrate, it is possible to achieve extremely high purity along with high quality SiC epitaxial layers. Activation annealing of the selectively doped area formed by ionimplantation at temperatures of more than 1600 C is among the challenges posed by the production of SiC devices. This has to do with the high stability of the material. Gate oxidation represents another challenge. Due to the remaining carbon clusters in the MOS interface (SiC + O2 => SiO2 + CO2 + CO + C), the channel mobility of SiC MOSFETs is very low compared to Si, leading to elevated channel resistances even at high gate voltages (Vgs) of i.e. 20V. Thus, the specific on resistance (RonA) of commercial MOSFETs is higher than the expected ideal values. Furthermore, this interface occasionally leads to unstable Vth values or poor Qbd values. Using a proprietary gate oxidation technology, Rohm managed to present its SiC MOSFETs to the market, featuring stable Vth values and high Qbd levels equivalent to Si MOSFETs. The device manufacturing process results in a so-called SiC-device wafer (Figure 4). In the subsequent processing steps, wafer is sawn and the devices are picked-up from this wafer for use in the final products (discrete packages or power modules). Figure 5: SiC SBDs feature better switching behavior than standard Si FRDs SiC SBDs feature much lower reverse recovery currents and shorter reverse recovery times, which reduces the relevant energy losses significantly. ROHM has introduced its technological advances into the market with its second-generation SiC SBDs. The cross section of a SiC SBD is depicted in Figure 6. Figure 6: Structure of Rohm s second-generation SiC SBDs Rohm diodes feature the lowest forward voltage worldwide (Figure 7). At the same time, they provide low leakage currents thanks to the precise manufacturing processes. Figure 4: SiC-device wafer from ROHM Semiconductor Figure 7: SiC SBD forward voltage at Tj = 125 C 6 Bodo s Power Systems January - March 2017 www.bodospower.com

Rohm s portfolio of second-generation SiC SBDs currently includes 650V products for 5A to 100A as well as 1200V and 1700V devices for a current level up to 50A. Automotive qualified SiC-SBDs from Rohm are also available. They are widely used in On-board charger Systems. Third-Generation SiC Diodes The motivation for the development of the third generation In applications like switched-mode power supplies (SMPS), SiC SBDs are now well known to be the better alternative to Si-based FRDs (Fast Recovery Diodes) in PFC stages (power factor correction). In this application, a large inrush current occurs at starting phase because the intermediate circuit capacitor (D.C. Link Capacitor) is not charged before turn-on. Due to the lower surge-current capability (IFSM) of Rohm s second-generation SiC SBDs, it is recommended to use bypass diodes in such SMPS applications. To support continuously downsizing requirements from the market Rohm designed its third-generation SiC SBDs meeting requirements of high surge current capability IFSM of the market (Figure 8). Initial products up to 10A are already available. implemented planar structure in their products. It is a well-known fact that a parasitic diode (the so-called body diode) is formed between the P layer and the N drift layer of a MOSFET (Fig. 10). In the history of SiC device development, so-called bipolar degradation has been one of the most critical issues. The device s on-resistance increases when a current is flowing in the body diode. Therefore, a stable behavior of the body diode is critical for the reliability of a SiC MOSFET in the final application. To ensure the reliability of their systems, power electronic engineers expect that the behavior of the body diode does not degrade. Figure 10: ROHM SiC MOSFET 2nd. Generation is based on a planar structure Figure 8: Structure of Rohm s new JBS SiC diode With the development of the Junction Barrier Schottky structure (JBS), Rohm managed to combine all advantages of SiC diodes in a single device. In this approach, P+ region are embedded underneath the Schottky barrier with optimum spacing in order to increase the diode s robustness while keeping the low Vf. But what influences the degradation of the body diode? Both crystal defects and manufacturing process of SiC MOSFETs have a great influence on the stability of the body diode. By acquiring the energy of hole-electron recombination when forward current flows, a certain type of a crystal dislocation changes its type from linear to planer shape. That can lead to a degradation of the on resistance of the body diode and the MOSFET. Based on its expertise in different manufacturing processes at the substrate, epitaxial growth and device level Rohm managed to prevent the degradation of the body diode. Figure 9: The new JBS diode combines the advantages of SBDs and PN diodes Due to the PN structure within the diode and the injection of minority carriers, the resistance of the epitaxial layer decreases with increasing temperature. On the other hand, the resistance of the epitaxial layer increases with the temperature of the SBD structure (Figure 9). SiC MOS Structures Rohm s Second-Generation Planar Structure The planar structure, which is among the most well-known structures of the semiconductor industry, also lends itself to SiC-MOS devices for high-voltage applications. Rohm is among leading suppliers that Figure 11: Rohm s SiC MOSFETs exhibit no on-resistance degradation Figures 11 and 12 illustrate the results of comparative measurements between Rohm MOSFETs and planar SiC MOSFETs from other manufacturer. In particular, 4 planar MOSFETs from other supplier were compared to 22 planar MOSFETs from Rohm. All the evaluated MOS- FETs feature a breakdown voltage of 1200V. Typical on resistance is 0.08Ω. Source current of 8A was conducted through body-diode. After 24 hours of continuous current flow, the on-resistance of the planar MOSFETs of other supplier had dramatically increased and blocking capability was lost. While Rohm s planar MOSFETs exhibited no performance degradation even after 1000 hours. www.bodospower.com January - March 2017 Bodo s Power Systems 7

Exclusively from Rohm: The Third-Generation Trench Structure For several decades, the trench gate structure has been a proven approach in low voltage Si-MOSFETs and Si IGBTs. This technology turned out to be beneficial for many power electronic applications. As the gate electrode is embedded into the drift layer, width of unit cell can be shrinked which enables higher current density. Exploring the conventional trench gate technology used for SiC MOSFETs, ROHM s designers found interesting facts. As an additional advantage of the trench structure, the on-resistance (Rdson) is reduced by 50% at the same die size, which contributes to a significant reduction of the conduction losses (Figure 14). Input capacity is also reduced by 35%, resulting in lower switching losses and a substantial reduction of the total energy losses. This structure is therefore considered to be an important step towards even more efficient modules featuring increased power density. Furthermore, the reliability of the gate oxide is improved by the reduction of the electrical field strength. ROHM started its volume production of thirdgeneration SiC MOS products with discrete SiC devices and full-sic modules based on its proprietary double trench technology. This expands the existing MOSFET product lineup and contributes to the design of highly efficient and highly reliable power electronics. Figure 12: Rohm s SiC MOSFETs exhibit no blocking voltage degradation As SiC features higher electrical field strengths than Si IGBTs, using the conventional trench gate structure would result in the following problem. In the off-state of a SiC device, a strong electrical field of approximately 2.66MV/cm occurs at the gate trench. The excessive stress applied to the gate oxide would degrade the reliability and lifetime of the devices significantly. Therefore, ROHM s double trench SiC MOSFET structure was designed to suppress this strong electrical field. In this approach, the source and gate electrodes are embedded into the drift layer (the source electrode is cut deeper than the gate electrode). As a result, the gate oxide is exposed to electrical field strength of less than 1.66MV/cm (Figure 13). Deeper source electrodes therefore prevent the concentration of electrical fields at the bottom of the gate. Figure 14: Trench SiC MOS devices feature 50% lower Rdson values Quality Assurance Measures for Rohm s SiC Devices SiC is a promising wide-band-gap material for industrial and automotive applications. Naturally, technology maturity and product quality are important factors for convincing the market that SiC can in fact meet the reliability and lifetime requirements of their systems. Quite often, however, it must still be determined how a semiconductor manufacturer like ROHM can ensure the quality of its SiC manufacturing process. Based on many years of experience in development and production of SiC and Si and big investments into its manufacturing sites, ROHM manages to meet and even exceed the reliability requirements. Possible Defects During SiC Wafer Production As reported in the first part of this article series and in numerous publications of research institutes and universities, SiC crystals can exhibit various defects including: Micro pipes Threading screw dislocation (TSD) Threading edge dislocation (TED) Dislocations in the planes perpendicular to the crystallographic main axis Figure 13: Comparison of electrical fields in single trench and double trench structures Most defects in the substrate result in damages to the layers during the epitaxial growth phase. On the other hand, other defects can also occur during the epitaxial growth phase, the ion implantation and dry etching processes. These spot defects usually emerge independent of the substrate s quality. 8 Bodo s Power Systems January - March 2017 www.bodospower.com

All kinds of macroscopic defects occurring during the epitaxial growth phase result in a significant increase of the leakage current and a degradation of the breakdown voltage which both influence the reliability of the SiC device. Figure 15: Quality assurance during Rohm s SiC manufacturing process For these reasons, it is essential to understand the physical properties of the materials used in order to retrace the defects that can occur during the manufacturing process. This enables a continuous improvement of the manufacturing process. Furthermore, Rohm conducts various tests during the manufacturing process in order to screen defective parts and to ensure full control over each processing step. This enables ROHM to ensure the delivery of sustainable products for high-volume markets. Rohm s Production Tests Rohm s quality control is based on 100% optical inspections and electrical tests. In addition, special inspections are made during the manufacturing process of SiC devices. SiC devices with visible defects generally fail during the electrical tests (gate-source or drainsource shorts). Nonetheless, Rohm makes an optical inspection at the beginning of device production to screen any substrate and epi-layer defects. In addition to visible defects, there may also be invisible faults including minor crystal defects in the substrate. This can even be more critical because devices featuring these invisible defects can operate flawlessly for an indefinite time but fail in the field, thereby degrading the reliability of the system. To prevent this, Rohm uses its unique screening technologies to detect invisible defects before delivery to the customer. The technical parameters of the devices are checked by electrical characterization at the end of the manufacturing process. For traceability reasons, all steps are documented for every single device. www.rohm.com/eu Aly Mashaly, Manager Power Systems Department Rohm Semiconductor GmbH Mineo Miura, SiC Power Device Engineer, Rohm Co., Ltd. Energy Efficient and Sustainable Future Systems with SiC ROHM Semiconductor, a leading enabler of SiC, has been focused on developing SiC for use as a material for nextgeneration power devices for years and has achieved lower power consumption and higher efficiency operation. Full Line-up SiC Wafer, SBD, MOSFET, Discrete and Modules Full Quality and Supply Chain Control In-House integrated manufacturing system from wafer processing to package manufacturing Leading Technology ROHM is the first semiconductor supplier worldwide who succeeded to provide SiC Trench MF Technology in mass production SMALLER STRONGER FASTER Full System Level Support Local system specialists provide comprehensive application support www.rohm.com/eu www.bodospower.com January - March 2017 Bodo s Power Systems 9

Increasing Power Density and Improving Efficiency While the creation of SiC substrate wafers was described in part 1 in BPS January 2017 and possible structures of SiC devices were the subject of part 2 in BPS February 2017, this third part will focus on the finished products. By Aly Mashaly, Manager Power Systems Department, Rohm Semiconductor GmbH Packages for SiC Semiconductors Basically, SiC diodes and MOSFETs can be used for all kinds of standard packages (discrete devices and modules). Discrete devices can be grouped into two categories: THT devices (through hole technology) and SMDs (surface-mounted devices). For package design, it is necessary to understand and consider the physical properties of SiC. For instance, parasitic inductances within the package are more significant in SiC semiconductors than in Si IGBTs because of the faster switching times of SiC devices. Therefore, Rohm makes intensive efforts to reduce the parasitic inductances within the discrete devices and modules. input and a galvanically isolated DC output. In practical applications, a Si MOSFET is used as the main switch with switching frequencies typically ranging between 16kHz and 500kHz. Higher switching frequencies enable the use of smaller inductors, although high losses in the Si MOSFET and the diode are a limiting factor. Rohm s new 1700V SiC MOSFET is excellently suited for these applications because of its efficient switching behavior and low Rdson. Rohm provides the 1700V SiC MOSFET in the TO-3PFM and the TO268-2L package (see Figure 1). Rohm began mass-producing of initial commercial SiC products in 2010. Since then, the company expanded its product lines across different power categories in order to meet the requirements of industrial and automotive applications. Rohm s product offering includes discrete devices (THT and SMD) as well as modules. As a semiconductor manufacturer, Rohm also provides its products as SiC wafers (bare die) and co-operates with major module manufacturers to expand the worldwide SiC market. Rohm s offering of second-generation SiC-SBDs currently includes 650V products from 5A to 100A as well as 1200V and 1700V versions for up to 50A. In 2016, Rohm started volume production of third-generation SiC-SBDs for 650V and currents up to 10A. In the SiC MOSFET segment, Rohm s product portfolio is even larger featuring two different technologies (planar and double trench). Planar technology is already available in discrete devices and modules for 650V, 1200V and 1700V and current levels up to 300A. Rohm starts its volume production of third-generation SiC MOS products with discrete devices and full-sic modules based on its proprietary double trench technology, which will expand the existing MOSFET product family and contribute to the evolution of highly efficient, highly reliable power electronics. New 1700V SiC MOSFETs for SMPS Applications Switch-mode power supplies (SMPS) are among the most important elements of all power electronic systems. After all, it is impossible to implement the functionality of the main system without a suitable power supply. Basically, the SMPS subsystem acts as the internal power supply for all electronic subassemblies of a power electronic system, providing the necessary supply voltages for the microcontroller, the sensors and the drivers of the main power semiconductors. As the industry s most popular SMPS topology, fly back converters of up to several hundred Watts transfer electrical energy between a DC Figure 1: Rohm s 1700V SiC MOSFET is available in two different package options Thanks to its high electrical field strength, the SiC MOSFET maintains its low on resistance per chip area even at high breakdown voltages. Figure 2 shows a comparison between Rohm s newly developed 1700 SiC MOSFET and the best in class Si MOSFET available in the market, which is typically, used in fly back applications. The 1700V SiC MOSFET features a much lower on-state resistance (Rdson) than the 1500V Si MOSFET. Values of 0.75Ω/1.15Ω have been achieved with the SiC MOSFET, while the Si MOSFET features 9Ω, although the SiC MOSFET s die size is 17-times smaller than the die size of the Si MOSFET. Using the 1700V SiC MOSFET (SCT2H12NY) in a TO268-2L package greatly improves the on-state resistance and the current capability compared to the Si MOSFET with the same package. Thanks to the SiC MOSFET s very low capacitance values (Ciss and Coss), switching frequencies beyond 100kHz can be used without any thermal problems, enabling a miniaturization of the inductive components and the space of printed circuit board. Furthermore, automated board assembly using a SMD line is enabled by the use of the TO268-2L SMD package, resulting in a significant reduction of production costs and finally total costs. In order to achieve optimum switching behavior of the SiC MOSFET in a flyback topology, Rohm designed a special 10 Bodo s Power Systems January - March 2017 www.bodospower.com

driver (BD768xFJ-LB) with a dedicated controller in a SOP-J8S package. In a three-phase 400VAC industrial application, two series-connected Si MOSFETs are typically used for flay back converter as the main switch to provide sufficient voltage headroom for the main switch in the off state. Thanks to the availability of Rohm s new 1700V SiC MOSFET, the two Si MOSFETs can be replaced by a single SiC MOSFET. Output filters of this kind enable the use of unshielded cables, which reduces total equipment costs dramatically. As an additional advantage, these filters lead to much lower high-frequency currents in the motor windings, which in turn results in reduced power losses in the motor. This, in consequence, will improve the motor s thermal behavior and reduce its noise level significantly. All in all, a sine filter will have a positive impact on the lifetime and reliability of the entire system. However, in high-power equipment the sine filter will account for most of the application s volume. It would therefore be an interesting proposition to reduce the filter s size and cost, which can be basically done by increasing the switching frequency of the semiconductors in the inverter. By default, Si IGBT technology is used in these applications at switching frequencies around 10kHz. IGBTs cannot be used in applications with elevated switching frequencies due to the high power losses and the resulting thermal stress in the inverter. With its excellent physical properties, SiC opens new doors to this kind of sophisticated applications. In particular, SiC enables the use of high switching frequencies without causing excessive thermal stress in the inverter. Two benefits result from using high switching frequencies: The output filter can be smaller and the resonance behavior of the entire system will improve accordingly. Figure 4 compares the switching losses of a Si IGBT technology and Rohm s SiC technology in a standard inverter topology (see Fig. 3). The comparison is based on a DC link voltage (Vdc) of 600V, a motor phase current (Imotor) of 200A and a switching frequency (Fsw) of 10kHz. Figure 2: Conduction behavior of Rohm s 1700V/1.15Ω SiC MOSFET compared to a 1500V/9Ω Si MOSFET High Efficiency in Drive Applications Industrial applications including drive inverters, which require sine filters at the output of the inverter, can benefit from the advantages provided by SiC. Examples include motor drive inverters whose power is in the double-digit kilowatt range. The motor and the inverter can be connected by cables of up to 100 meters in length. A sine filter is often used at the output of the inverter in these applications. Figure 4: Comparison between a Si IGBT, a hybrid configuration and a Rohm full-sic module For the data in Figure 4, the loss was normalized to the combination of Si IGBT and Si FRD (black bar). In a hybrid configuration consisting of a Si IGBT and a SiC SBD as an antiparallel freewheeling diode, switching losses will be 30% lower (blue bar). If a full-sic module from Rohm (SiC MOSFET and SiC SBD) is used in this topology, switching losses will be reduced significantly (red bar). The world s first full SiC module based on Rohm s newly developed trench technology is shown in Figure 5. Featuring a breakdown volt- Figure 3: Standard topology of a drive inverter Figure 5: Full-SiC module based on Rohm s trench technology www.bodospower.com January - March 2017 Bodo s Power Systems 11

age of 1200V, the module supports a drain current of 180A. Rohm is increasing its line up from Modules continuously and offers now modules up to 600A. Reverse Conduction of SiC As described in part 2 of this article series, a parasitic inverse PN diode (body diode) is present in every SiC MOSFET. In commutation loops of power electronic systems with series inductors, some current flows in this body diode when the SiC MOSFET is in the off state. Like all PN diodes, a forward voltage occurs at the diode when a current is flowing through it. Physical properties of SiC MOSFETs include a wide band gap, which results in a high voltage drop at the body diode. During the design phase of a system, power electronic designers should explore how often the body diode will be used during switching transitions. The high voltage drop results in high thermal losses if a large share of the current flows through the body diode. In addition, switching time plays an important role in thermal design. As one of the options available to eliminate this high voltage drop, it is possible to benefit from a specific physical property of SiC MOSFETs called reverse conduction (see Figure 6). The mechanism behind reverse conduction is based on the fact that the channel of a SiC MOS- FET can be turned on again even if the MOSFET is reverse biased. The reverse current will then flow through the SiC MOSFET channel instead of the body diode. As depicted in Figure 7, the voltage drop between drain and source will decrease if the gate voltage increases. A gate voltage of Vgs = 18V will yield optimum results. Figure 6: Reverse conduction of SiC How can this reverse conduction effect be used and adjusted in applications based on half-bridge configurations? The half-bridge configuration is a very popular topology in power systems including inverters and DC/DC converters. It consists of two switches or SiC MOSFETs (high-side switch and low-side switch). In the half-bridge topology, the complementary driver signals generated by the microcontrollers are adapted to the MOSFETs by gate drivers to ensure the functionality of the system. A dead time must be inserted between the transitions of both switches to prevent a short circuit in the DC link. The current will flow through the body diode during the dead time. After the dead time, the MOSFET, which previously was in the off state, can be turned on again although it is reversebiased. As shown in Figure 8, the current will flow through the body diode only during the dead time (blue arrow) before the MOSFET channel is turned on again to let the current flow though the MOSFET channel (red arrow). This reverse conduction can be implemented via the software running on the microcontroller. Figure 8: Using reverse conduction in the half-bridge topology About Rohm Semiconductor Rohm Semiconductor is a leading supplier of systems, LSIs, discrete devices and modules. Using latest semiconductor process technologies and state-of-the-art automation technology for proprietary manufacturing systems, Rohm can maintain its leading position through high operative expertise and outstanding product quality. In 2015, the Rohm Group recorded sales of 3 billion dollars with 20,843 employees. Rohm provides a wide portfolio of IEC-Q-qualified products for the automotive industry, ranging from standard products including transistors, diodes, EEPROMs, operational amplifiers, comparators and LDOs to ASIC and ASSP products including LED drivers, motor drivers and optimized gate drivers for engine control units (ECUs). Furthermore, Rohm provides leading technologies for the new generation of power products, resulting in breakthrough performance with respect to efficiency, heat generation and compactness. New products and materials like SiC MOSFETs and full-sic modules were developed for this purpose. Rohm makes unmatched contributions with its energy-efficient, futureoriented solutions especially tailored to meet the requirements of the automotive market. www.rohm.com/eu Figure 7: The voltage drop will decrease when the MOSFET channel is turned on again 12 Bodo s Power Systems January - March 2017 www.bodospower.com