source：Huafeng Capital
2.2 GaN industry chain: a self-contained system from materials to devices, upstream controls the development direction
The GaN industry chain is self-contained, and the upstream material R&D and manufacturing segment determines the development direction and speed of the entire industry, which is the most significant difference from the Si industry chain. The GaN industry chain includes upstream substrate manufacturers, epitaxial wafer manufacturers, midstream GaN device design, manufacturing, and packaging manufacturers, and downstream application companies. Since the preparation of GaN and the size of wafers that can be achieved still have a large gap compared with Si, the overall development of the current industry chain is still controlled by the preparation of upstream materials.
source：Huafeng Capital
As far as the entire GaN industry is concerned, mainstream international companies adopt the IDM model, and domestic companies have a large gap in technology, production capacity, and customer trust. GaN international manufacturers include Qorvo, Sumitomo, NXP, Mitsubishi, etc., all of which are IDM models, and their production capacity, solutions, market channels, and customer trust are all perfect. Domestic IDM manufacturers mainly include CLP 13/55, Suzhou Nengxun, Suzhou Nenghua, etc., but they are far behind the mainstream international manufacturers.
IDM manufacturers in the GaN industry have a development trend toward foundry business. As an industry pioneer and a leader in GaN-on-SiC MMIC technology, Wolfspeed under Cree uses the world's largest wide bandgap semiconductor production line to provide customers with design assistance to manufacturing and testing services, shortening the product launch cycle for downstream customers. Traditional silicon wafer foundry TSMC has begun to provide GaN foundry services, and Sanan Optoelectronics is also doing GaN foundry projects.
source：Huafeng Capital
2.3 Heterogeneous substrates and epitaxy: each has its own expertise, GaNonGaN is the ultimate choice
The type of substrate used for GaN epitaxial wafers directly determines the quality of GaN epitaxial wafers, and has a decisive influence on the overall performance that GaN devices can achieve and their downstream application prospects. GaN devices can be divided into many types according to different growth substrates. The degree of adaptation of different substrate materials to GaN is shown in the figure below. The sapphire substrate has a high degree of mismatch with the GaN lattice and has poor electrical performance, so it is difficult to make power electronics and radio frequency devices, but it is widely used in the LED field because of the better optical performance of sapphire.
source：Huafeng Capital
GaNonSi has a prominent price advantage. Based on the mature process system of existing Si materials, the device has the most advantage in the competition of mid-power devices, and it is expected to increase volume quickly. GaNonSi has low cost and fast working speed, making it an ideal material for medium-power and high-efficiency power electronic devices. In the high-power field, due to the poor lattice fit and thermal conductivity of Si, device performance is difficult to benchmark against SiC devices and GaNonSiC devices. GaNonSi currently used for power electronic devices has basically achieved 6-inch mass production and 8-inch R&D, and it is expected to expand market capacity soon.
GaNonSiC has outstanding efficacy in electronic power devices, and the future in the high-power field can be expected. SiC itself has the highest thermal conductivity, the best power performance, and the strongest heat dissipation capacity of the device. Combined with GaN's own excellent switching characteristics and high-frequency characteristics, it is expected to obtain the best application scenarios in the high-power field.
The overall performance of GaNonGaN is the most outstanding, but the preparation technology is the most difficult, and it is prioritized to be used in the LED and laser fields before the cost drops. However, benefiting from the lowest lattice adaptation of epitaxial wafers grown on GaN substrate materials, the manufactured devices are expected to completely replace the existing sapphire-based GaN and Si-based GaN solutions in the fields of RF, LED, and power electronics. And will have their own strengths with GaNonSiC, becoming the ultimate choice of the GaN industry chain.
2.4 GaN single crystal preparation-the core of the industry chain: technology diversion Japanese companies are in a monopolistic situation
The performance of GaN epitaxial wafers is the priority for the development of GaN materials. GaN devices have excellent performance, and the substrate material and its epitaxial wafer are the foundation. Therefore, the development of GaN devices is applied to various fields with the principle of giving priority to performance and taking into account the cost. Because the GaN semi-insulating substrate used in homoepitaxial GaN-on-GaN is very difficult to manufacture, and the existing semi-insulating GaN substrates in the world are almost all 2 inches, currently mainly heteroepitaxial, such as GaN-on-sapphire, GaN-on-Si, GaN-on-SiC, etc. However, the inherent shortcomings of heteroepitaxial technology, such as lattice mismatch, thermal expansion mismatch, and poor chemical mutual solubility, lead to high dislocation density, mosaic crystal structure, biaxial stress, and warpage of epitaxial wafers in the GaN substrate prepared by epitaxy. And other defects.
The heteroepitaxial technology has a high dislocation density, and it is difficult to guarantee the power and stability of the manufactured GaN device. GaN epitaxial wafers produced by heteroepitaxial technology, on the one hand, due to the high dislocation density, will reduce carrier mobility, luminescence lifetime and material thermal conductivity, and dislocations will form non-radiative recombination centers and light scattering centers. Therefore, the luminous efficiency of the electronic device is reduced; on the other hand, since the electrode metal and impurity metal elements diffuse into the dislocations, the leakage current formed will reduce the output power of the device and seriously affect the stability of the device.
Low defect density, thick film and large size are the direction of substrate development. At present, the development of GaN devices is advancing with the logic of giving priority to performance and taking into account the cost. For example, 5G base station PA uses GaN-on-SiC instead of GaN-on-Si. Defect density is an important factor affecting the performance of GaN devices, so low defect density is an important development direction for GaN substrate materials. In terms of large size, since GaN substrates are currently mainly 2 inches, 4 inches are few, and 6 inches are basically not, the application of GaN substrates on devices is greatly restricted.
source：Huafeng Capital
The HVPE method currently accounts for 85% of the GaN substrate market. Japan is currently the largest player in the HVPE law. The layout companies include Mitsubishi Chemical, Hitachi Cable and Sumitomo Chemical. Among them, Sumitomo Chemical can currently produce a small amount of 4-inch semi-insulating GaN substrates, but it is temporarily unable to mass-produce due to process technology issues.
Domestically, Dongguan Zhongjia can produce a small amount of 4-inch products under laboratory conditions, while Suzhou Nawei and Shanghai Gaite currently only have 2-inch products. The three manufacturers all use sapphire substrates and HVPE methods to produce GaN substrates, all using the lift-off method. In general, in 2019, my country's homogeneous substrates have been industrialized in small batches of 2-3 inch substrates, and 4-inch samples have been realized, and the comprehensive indicators have reached the domestic advanced level.
The HVPE (Hydride Vapor Phase Epitaxy) method is currently the most widely used GaN substrate growth method, and only the seed crystal SCAM with the lowest lattice mismatch can exert its maximum effect. The HVPE method has simple equipment, fast growth rate, and large-size uniform growth. The problem is that the HVPE method using a heterogeneous substrate has a very high defect density, which greatly affects the performance of the device, and it is difficult to prepare large-sized GaN seed crystals. At present, the industry's pioneering approach uses R acid (CA) as the seed material. The lattice mismatch with GaN is only 1.8%, and crystal defects can be suppressed very low.
source：Huafeng Capital
The HVPE law is the best path for industrialization, and the current technology is mainly monopolized by Japanese companies. The ammonia heating method and sodium melting method are only in the laboratory stage, the process conditions are harsh, the equipment requirements are high, and the size of the grown GaN single crystal is small. In particular, in order to grow large-size GaN, the ammonia heating method requires a larger amount of gas to be added, and the safety is difficult to guarantee. HVPE (Hydride Vapor Phase Epitaxy) technology has become the mainstream technology for mass production of GaN substrates due to its high growth rate, low process cost, and simple equipment maintenance. Although Japan's Mitsubishi Chemical has already deployed the ammonia thermal method, it uses the ammonia thermal method to produce small-size GaN and then the HVPE method for further production. The famous Polish ammonothermal method company Ammono is limited by ammonothermal method and is difficult to mass produce, and is currently on the verge of bankruptcy.
The current mainstream product of GaN self-supporting substrate is 2 inches, and 4 inches has been commercialized. The main production process of GaN self-supporting substrates is the lift-off method. If the crystal column cutting method achieves a breakthrough, it will greatly reduce the production cost of GaN devices and accelerate the large-scale application of GaN devices.
The cutting method is an ideal process to reduce costs. From the perspective of dismantling GaN production costs, crystal growth accounts for 30%, and subsequent processing accounts for as much as 70%. Generally, the production of GaN substrates uses the lift-off method. The lift-off method refers to the method of growing a thin GaN substrate on sapphire and then peeling it off. The process is relatively complicated and the production efficiency is not high. The way to improve production efficiency should be explored from the root cause of improving GaN growth ability. Analogous to silicon-based semiconductors: silicon growth is easier. The general single crystal silicon growth path is to first prepare polycrystalline silicon or amorphous silicon, and then use the Czochralski method or suspension zone melting method to grow rod-shaped single crystal silicon from the melt. Then cut thin wafers one by one. Therefore, if the GaN substrate can be made with the same long crystal column and cut into thin slices, its production efficiency may be improved by several orders of magnitude. At present, domestic frontier companies have made arrangements in the cut-off method.
2.5 GaN application fields and market scale
2.5.1 RF applications: 5G strong winds boost GaN to fully enter the market. Domestic manufacturers show a latecomer and catch-up trend
The demand for higher data transmission rates in the telecommunications industry and the demand for higher resolution in industrial systems have contributed to the continuous increase in the operating frequency of electronic devices used in these fields. At the same time, many systems need to work in a wider frequency spectrum, making new designs usually require increased bandwidth. In this context, people tend to use a signal chain that covers all frequency bands. The emergence of GaN revolutionized the entire industry.
Compared with different semiconductor process technologies, SiGe is only suitable for lower working voltages (2V-3V), but it has high compatibility with silicon integrated circuit processes, so there are still applicable scenarios. GaAs is suitable for microwave frequencies with a working voltage of 5V-7V and has been widely used in power amplifiers for many years. The silicon-based LDMOS technology has a working voltage of 28V, but its device power will decrease as the bandwidth increases, so it is difficult to adapt to application scenarios above 3.5GHz.
source:analog
source:analog
LDMOS is not suitable for high frequency, GaAs needs multi-level superposition to provide sufficient power, and GaN's high temperature, high frequency, and high power characteristics make it prominent in RF components. At present, GaN has become the main rival of silicon-based LDMOS devices and GaAs in the high-frequency field in RF power applications. Silicon-based LDMOS devices can only be used for applications below 3.5GHz, and although GaAs can be applied to the high-frequency field of 40GHz, it requires multi-stage amplification and stacking to achieve the power index, so the device size is usually relatively large. However, GaN can not only adapt to high frequency bands, but also provide high power characteristics, which can greatly reduce the number of transistors and device size. The figure below shows a comparison between a multi-stage GaAs power amplifier and its equivalent GaN power amplifier. It can be seen that compared with GaAs, the chip size can be reduced several times under the equivalent power using GaN.
source:analog
The band gap voltage of GaN is higher than that of silicon-based LDMOS devices and GaAs, which can adapt to higher voltages. The band gap voltage is determined by the characteristics of the semiconductor material itself, and directly determines the highest voltage that the manufactured device can adapt. The bandgap voltage of GaN is 3.4eV. Compared with Si bandgap voltage of 1.12eV and GaAs's 1.42eV, the characteristics of deep bandgap make GaN devices have a higher breakdown voltage, which allows the use of higher rails to meet Higher power requirements. In comparison, the power rail of GaN is usually 28-50V, and GaAs is 5-7V. For devices of the same size, GaN devices have higher power density. At the same time, based on the good heat dissipation performance of GaN, its devices can reduce power consumption better than GaAs.
Source: Microwave Magazine
The key technology to improve 5G system capacity and spectrum utilization is Massive MIMO, which has a demand for small-sized, highly integrated power amplifiers. MassiveMIMO technology refers to a large-scale antenna array technology, which uses multiple transmitting antennas and multiple receiving antennas to improve the utilization rate of the spectrum, thereby realizing space diversity. 4GLTE has achieved through OFDM (Orthogonal Frequency Multiplexing, an excellent multi-carrier modulation technology) to solve the problem of selective deep fading of its frequency, but generally only supports 2 antennas, 4 antennas, the maximum number of antennas for uplink and downlink diversity Limited to 8 or less. The number of antennas and channels of MassiveMIMO technology under 5G can reach 64/128/192. This makes 5GMassiveMIMO compared with 4GLTE, the digital signal processing performance is improved 16 times. The challenge of this kind of system design puts forward new demands on small size, high performance, low cost PA. In addition, because of the increased complexity of the 5G modulation mechanism, PA needs to be very efficient under deep back-off conditions (up to or more than 8dB).
Source: Antenna Industry Alliance
GaN-on-Si makes the MassiveMIMO antenna array of Sub-6GHz5G macro base station highly integrated, while meeting high frequency and bandwidth characteristics. PAs deployed by OEMs in 5G base stations and networks below 6 GHz are transitioning from silicon-based LDMOS technology to GaN. Compared with silicon-based LDMOS, GaN itself can be used in the band above 3.5GHz. Its high breakdown voltage, high current density, and low on-state resistance make GaN have high efficiency under high frequency band and wide bandwidth conditions, making MassiveMIMO systems capable of More compact. At the same time, GaN can operate reliably at higher temperatures, so a smaller heat sink can be used to achieve a more compact size.
The RFEE module of 5G millimeter wave micro base station has higher frequency, higher bandwidth, higher data rate and lower system power consumption requirements. RFEE is only critical to the power output, selectivity and power consumption of the entire radio frequency system. Under 5G, RFEE has requirements for integration, downsizing, lower power consumption, high output power, wider bandwidth, improved linearity, and increased sensitivity. In addition, there are also requirements for coupling with transceivers and antennas.
The introduction of GaN-on-SiC PA enables the dense integration and functions of 5GRFEE to be realized, and at the same time saves space for "inches of land" PCB boards. Its specific advantages are: (1) Provide higher frequency and higher bandwidth: DohetryPA using GaN-on-SiC has wider bandwidth and higher power added efficiency than silicon-based LDMOS devices. (2) Provide a higher speed, benefit from the soft compression characteristics and cross-band characteristics of GaN. (3) Minimize system power consumption: GaN greatly reduces system power consumption while providing high output power.
Source: Semiconductor Industry Observation
The amount and rate of 5G mobile phone radio frequency modules are rising, and GaN may become the only choice for mobile phone PA in high frequency bands. Generally, 3G and below mobile phones are equipped with 1-2 PAs. The average number of PAs in 4G mobile phones is 3-6. The application of Massive MIMO technology under 5G will greatly increase the antennas of the mobile phone. Correspondingly, the number of PAs in a single mobile phone will also increase. Ushered in significant growth, it is expected to reach 16 or more. At present, Sub 6GHz mobile phone RF devices are still the main field of GaAs, but after the mid-to-long-term development to the higher frequency millimeter wave stage, GaAs has low thermal conductivity and poor heat dissipation, and its RF devices can withstand relatively low power. Can not be applied to mobile phone PA above 28GHz, GaN is the only choice.
The introduction of GaN power devices in smartphones needs to solve the problems of system, mobile phone operating voltage and manufacturing cost, and the landing time will not be fast. On the smartphone side, GaN applications also face some problems: (1) The linearity optimization of GaN power devices needs to be combined with digital technology (2) The operating voltage of smartphones needs to be increased to fully develop the GaN efficiency bandwidth advantage (3) The cost needs to be further reduce.
GaNonSiC is difficult to use in millimeter-wave-level frequency bands, and GanonGaN will become the ultimate solution for 5G. SiC currently benefits from better thermal conductivity, and the combination with GaN can play a certain role. However, when applied to the millimeter wave level of the 5G frequency band, devices made of SiC substrates are difficult to benchmark against devices made of GaN self-supporting substrates, and their output power and operating frequency are difficult to reach the required standards for millimeter waves.
The development of domestic GaN radio frequency devices is catching up with the international level. In 2019, the working frequency band of high-power GaN devices has reached the Ku band, and the output power has reached the kilowatt level. The working frequency of GaN microwave power monolithic integrated circuits reaches the W-band, and the output power exceeds the order of hundreds of watts; Sub-6GHzGaN power devices used in 5G mobile communications have realized serialized products below 600W, and have launched lines based on 0.5µm and 0.35µm Of mass-produced products. The performance of silicon-based GaN radio frequency devices is at the forefront level in the world, and the operating frequency is 145GHz-220GHz. Large-scale and quantitative supply has been achieved. After installation verification, the one-time installation through rate is less than 300ppm.
Source：Yole
Chinese manufacturers in the GaNRF field as a whole belong to the middle and lower reaches, and there is a lot of room for development. In the GaNRF device global manufacturer strength map produced by Yole, CLP belongs to the international midstream position, and the new Hiwei Huaxin and Huajin Chuangwei are new entrants in the industry. On the whole, there is a significant gap between domestic manufacturers and international leading manufacturers Cree, Toshiba, and Fuji.
Market: Bright and quantifiable development prospects
According to Yole data, the global GaN radio frequency device market reached 537 million U.S. dollars in 2019 and is expected to reach a market size of 1.324 billion U.S. dollars in 2023.
Source：Yole
5G base station construction is currently the largest market for GaN. According to Yole's statistics, the market size of GaN radio frequency devices for base stations reached US$150 million in 2018. 5G commercial macro base stations will be based on MassiveMIMO array antennas with up to 64 channels. According to a base station with 3 sectors, the demand for a single base station PA module is as high as 192.
Source：Yole
It is estimated that by 2023, the penetration rate of GaN PA modules in base stations will exceed 85%, driving the demand for GaN PAs to exceed 194 million. It is estimated that by 2023, the market size of GaN radio frequency devices for base stations will reach 521 million U.S. dollars, with a CAGR of 28% in 2018-2023. The market segmentation by application area is as follows:
GaN's defense market is expected to grow from US$270 million in 2018 to US$977 million in 2024, with a CAGR of 23.91%.
The market size of GaN wireless infrastructure will grow from US$304 million in 2018 to US$752 million in 2024, with a CAGR of 16.3%.
The GaN wired broadband market has grown from USD 15.5 million in 2018 to USD 65 million in 2024, with a CAGR of 26.99%.
The scale of the GaN RF power field has grown from US$2 million in 2018 to US$104.6 million in 2024, with a CAGR of 93.38%, which has a lot of room for growth.
Domestically, the scale of my country's GaN microwave radio frequency market in 2019 is approximately 4.856 billion yuan, a year-on-year increase of 200% compared to 218. It is estimated that from 2019 to 2024, the market is expected to maintain an annual growth rate of 42%.
Sources of GaN RF market classification and share in 2019：CASAresearch
In 2019, my country's 5GGaN radio frequency application scale reached 2.3 billion yuan, which is the main driving factor for the GaN radio frequency market. According to data from China Academy of Information and Communications Technology, my country's demand for 5G macro base stations in 2020 will reach nearly 500,000.
As far as the entire industry is concerned, the output value of my country's GaN radio frequency industry has grown rapidly in recent years, from 278.5 million yuan in 2016 to 3.8 billion yuan in 2019, and the volume has increased nearly 15 times in four years. In terms of breakdown, the GaN device field is the largest part of the RF market, accounting for more than 60% in 2019. It is expected that with the maturity of the substrate technology and the further improvement of the overall performance of the device, my country's GaN radio frequency market will maintain sustained rapid growth.
Source：CASAresearch
2.5.2 Electronic power field: replacing Si-based devices, the first in the field of fast charging
In the field of power electronics, compared with traditional silicon semiconductor materials, it allows power devices to operate at higher voltages, frequencies, and temperatures. Its advantages are mainly reflected in:
(1) The band gap of GaN is 3 times that of Si, and the breakdown electric field is 10 times that of Si. Therefore, the on-resistance of a GaN switching power device with the same rated voltage is 3 orders of magnitude lower than that of a Si device, so the conduction loss is greatly reduced.
(2) Since the GaN transistor does not contain a body diode, there is no reverse recovery loss.
(3) The input charge of the GaN transistor is very small, and there is no gate drive loss.
(4) The power density of GaN devices is very large, more than four times that of silicon-based LDMOS devices.
(5) GaN power devices support a higher switching frequency (1MHz, compared to Si is <100KHz), which can greatly reduce the volume of passive devices, making GaN devices very attractive.
Source: Huagong Semiconductor official website
The application of GaN in power electronic devices is mainly power equipment, such as power adapters, wireless charging equipment, etc. Compared with SiC, GaN is more suitable for 100-600V low and medium voltage applications. The reason is that GaN devices are mostly lateral structure (JEFT), and it is difficult to achieve the high voltage capability of SiCMOSFET (vertical structure). Because the structure contains heterojunction two-dimensional electron gas that can achieve high-speed performance, GaN devices can have higher frequency performance than SiC devices. And because it can carry voltage lower than SiC devices, GaN devices are mainly suitable for power devices with high frequency, small size, cost requirements, and relatively low power requirements.
Specifically, GaN is expected to gain favor in areas such as automotive electronics and fast charging.
Fast charging: GaN fast charging has become a hot spot in the industry. Core GaN chips are still domestically blank
The charger using GaN has the advantages of small size (only 1/4 of the original), light weight, high conversion efficiency, low heat generation, and strong safety. Depending on the internal circuit architecture of the charger, 1 or 2 GaN power electronic devices will be used, with an average conversion efficiency of over 90%. According to Gartner data, the global average annual shipment of smart devices exceeds 2 billion units. With the rapid increase in GaN market penetration, the fast charging market is expected to become the biggest driving force for GaN in the power electronics field.
The core principle of GaN fast charging chargers to reduce the size is that GaN's excellent switching frequency improves the efficiency of high-frequency transformers and reduces the size. A core requirement of the charger is to convert the 220V voltage commonly used in our country to the maximum voltage that electronic products can withstand (for example, iPhone products are generally 5V). It is a high-frequency transformer that realizes this function, which occupies a relatively large space in the charging head. Increasing the switching frequency of the transformer can increase the number of energy conversions per unit time, thereby increasing the conversion power. Compared with silicon-based power devices, the switching speed of GaN power devices has increased by more than 100 times, which allows the volume of high-frequency transformers to be greatly reduced, thereby greatly reducing the volume of the entire charger.
High-frequency transformer for charger Source: Bafang Resources
In 2018, Navitas and Exagan, the world's first GaNIC manufacturers, launched a 45W fast charging power adapter using GaN integrated solutions. By the end of 2019, many manufacturers have launched GaN fast chargers. Domestic manufacturers include OPPO and Xiaomi. At the Huawei P40 conference in 2020, Huawei even released a 65WGaN dual-port super fast charging charger. The product supports Type-A and Type-C dual-port charging.
Source: public data compilation
Source: Zimi official website Source: charging head
The figure above shows the volume comparison chart of Zimi brand GaN65w power adapter and Apple usb-C61w power adapter. It can be seen that the cross-sectional area of the GaN fast-charge charger is 50mmX50mm, which is more than 45% smaller than the charger for Apple's ipad, and the volume is reduced by nearly 50%.
Source: "Summary of GaNFET Structure, Driver and Application"
Disassembling the ANKER GaN charger, the "core" for reducing the volume and increasing the power density is the NV6115 and NV6117GaN chips produced by Navitas. This monolithic GaNFast power integrated circuit has a rated power of 650V and can be used in a 5×6mm QFN package to achieve high-speed, high-frequency, and high-efficiency operation in many soft-switching topology applications. There is currently no similar product in China.
Automotive electronics: rich application scenarios
GaN has various application scenarios in the field of automotive electronics. For the new high voltage of the car (such as 48V)
For bus systems, GaN can improve efficiency, reduce size and reduce system costs. At the same time, Lidar uses pulsed lasers to quickly provide high-resolution 360 three-dimensional images of the surrounding environment of the vehicle. GaN technology can make the laser signal transmission speed much higher than similar silicon devices, enabling autonomous vehicles to see farther, faster and clearer . GaN has high working efficiency and can achieve maximum wireless power system efficiency at low cost. When used in high-intensity LED headlights, GaN technology can increase efficiency, improve thermal management and reduce system costs. The higher switching frequency allows operation above the AM band and reduces EMI. On the whole, GaN has rich application scenarios in automotive electronics.
GaN transistors are expected to become the first choice for inverters in the mid to late 2020s. In 2019, many companies' GaNHEMT products have obtained JEDEC certification in the semiconductor industry, and have successively obtained automotive-grade AEC-Q101 certification, which has strengthened users' confidence in the reliability of GaN transistors. According to HIS Markit analysis, GaN transistors may be the first to break through the bottleneck of large-size epitaxy, thereby reducing prices. Compared with SiC MOSFETs, GaN transistors may become the first choice for inverters in the mid to late 2020s.。
Source：Yole
Market: Domestic GaN power device localization is extremely low, development needs upstream promotion
Based on Yole and HIS Marikt's forecast, the market size of GaN electronic power devices in 2019 will be approximately US$76 million. It is estimated that by 2024, the global GaN power electronics market is expected to exceed US$600 million, with a CAGR of 41%.
Source：Yole
Dividing the GaN market share by voltage size, GaN's most advantageous voltage range is 300-600V, and SiC devices are generally used in higher voltage ranges. Currently, 600/650V products have the largest market scale in the power electronics field, and are expected to account for 80% of the entire GaN power electronics market, while 900/1200V devices and 200V devices will each account for 10%.
At present, the highest voltage of Si-based GaN electronic power devices that have been commercially promoted is still 650V, but the room temperature current has reached 150A, which is a 25% increase in performance compared to 2018. According to CASAresearch, the withstand voltage level of GaN-on-Si power electronic devices is expected to increase to 1200V in the next 5-10 years. Transphorm's 900VGaNFET devices have begun trial production. Domestically, the commercial Si-based GaNHEMT has a working current of 35A, but the research and development progress has reached the domestic advanced level.
Source: "The Third Generation Semiconductor Industry Report 2019"
From a domestic perspective, based on the comprehensive data of SiC and GaN in the power electronics market, the power supply sector is the largest market with approximately 1.62 billion yuan, accounting for 58% of the entire third-generation semiconductor power electronic device market.
In terms of power, the low-voltage segment accounts for 68% of the market for all application voltages of GaN power devices, and is expected to continue to expand in the future.
Source：Yole
According to CASAresearch data, my country's power semiconductor market has a low degree of localization, and more than 90% of IGBTs are imported. The penetration rate of GaN in the power electronics field is less than 1%, and more than 90% are also dependent on imports. The main manufacturers are Cree, Infineon and Rohm.
The low level of localization in the power semiconductor market needs to be resolved from the breakthroughs in upstream GaN substrates, epitaxial manufacturers' technology and mass production capabilities.
2.6 LED field: GaN can achieve full spectrum coverage, outstanding advantages in laser field
LED is the most applied field of GaN-on-Sapphire, and laser is considered to be the first field where GaN self-supporting substrates will be applied. GaN blue LED has been launched as early as 1993, which solved the problem of the lack of high-efficiency blue light in the LED field. In 1996, the white light emission of LED was realized.
GaN-on-Si is the natural choice for MicroLED chips. MicroLED specifically refers to the self-luminous LED whose size is 3-10μm. Its current main potential market is the high-resolution home consumer electronics market. Depending on the final application scenario, MicroLED can directly produce high-resolution displays on Si, GaN or sapphire and on the ground, or after the substrate manufacturer is completed, the chip can be made in a larger size and with logic through a massive transfer method. The circuit board is assembled.
The solid solution formed by GaN and other nitrides can cover the spectrum from infrared to ultraviolet. It can be mixed with InN and AlN in any ratio to form a solid solution. The forbidden band width of these solid solutions can change continuously between 0.7eV and 6.2eV, so it can cover the infrared to ultraviolet spectral region, and because these nitrides are all Direct band gap semiconductors, so they have received extensive attention in the visible and ultraviolet light emission side.
Energy-saving and environmentally friendly GaN-based LEDs are widely used in solid-state lighting, mobile phone and computer display screens, and traffic signals. GaN-based semiconductor lasers can be used in the field of optical storage. Because the wavelength is shorter than that of GaAs-based semiconductor lasers, GaN-based semiconductor lasers can effectively improve storage density and read/write speed.
Laser light sources based on semi-polar GaN laser tubes have outstanding advantages over blue LEDs and have been applied in specific lighting fields. In 2014, BMW launched the world's first automotive laser headlights. Audi has also launched laser car light products on its high-end products. Its laser headlights have an illuminating distance of more than 600 meters, which is more than double the illuminating distance of LED lights.
Source：BMW
my country's technological research progress in the LED field has ranked among the world's top levels. In 2019, the power GaN-LED luminous efficiency exceeded 2001m/W. The industrialized luminous efficiency of power silicon-based LED chips reached 1701m/W, and the luminous efficiency of yellow LEDs was greatly increased to 27.9%, creating the highest international value. In terms of new display technology, the wavelength consistency, epitaxial size and yield of the small-pitch display and MiniLED blue-green display chips (50-100μm) have been greatly improved to meet the requirements of industrialization; in the field of ultraviolet LED, China’s industrialized UVA LED Chip (390nm-400nm), luminous power reaches 980mW -1060mW@500mA. In terms of curing applications, 385nm and 395nm UVA device technologies have realized the replacement of original light sources in ink curing equipment; in the field of sterilization and disinfection, the output power of 280nm devices can meet the needs of low-power disinfection products.
2.7 Challenge: The failure mechanism and reliability evaluation system of materials, devices, and packaging have a long way to go
As the GaNonSi material that will be applied first in the GaN system, the problems it faces are relatively simple, but the most direct, so we will make a brief analysis of its application problems.
(1) GaN materials on large-size Si substrates are easy to crack and have high defect density, which makes it difficult to support the application environment of 1200V devices.
(2) It is an international problem during the preparation of high-pressure resistant, stable and reliable normally-off HEMT.
(3) High frequency and high power density place extremely high requirements on GaN device packaging.
(4) Drive and electromagnetic compatibility also require the same degree of miniaturization.
From this point of view, the problems faced by GaN devices do not entirely stem from the quality of the substrate and epitaxial wafers, but must be considered in the development process of the downstream segment. GaN industry ecological construction and technological development have a long way to go.
3. SiC
3.1 Introduction to SiC: There are various structures, and the three main forms of SiC have their own applications
SiC is the only stable compound of Si and C. SiC crystal is one of the hardest materials known, and its hardness is as high as 9.2-9.3 on the Mohs at 20°C. The sublimation temperature of SiC is as high as 2300℃.。
Source: China Knowledge Network, Zhihu
At present, there are more than 200 different crystal structure forms of silicon carbide, but only one allotrope belongs to the cubic crystal system, namely 3C-SiC, also known as β-SiC. SiC with different crystal structures has different applications. Among them, the α crystal type 4H can be used to manufacture high-power devices, and the 6H crystal type as the most stable crystal type can be used to manufacture optoelectronic devices. The β-SiC crystal type can be used to manufacture high-frequency devices and other thin film materials (such as GaN) substrates due to its special zinc blende structure.
3.2 SiC industry chain: the international mainstream is IDM manufacturers, and domestic companies start from the upstream layout
Source: Huafeng Capital
In the future, the core of the development of domestic SiC manufacturers will start with the quality and size of upstream SiC materials. At present, the development models of leading international companies are mostly in the form of IDM manufacturers, such as Cree, Toshiba, and Sumitomo. The domestic IDM manufacturers mainly include CLP and Nengxun Semiconductor. The SiC materials currently produced by international manufacturers still have a large gap with the international advanced level in terms of substrate and epitaxial size and quality, which directly leads to the lagging development of SiC design and packaging manufacturers. In the future, domestic manufacturers are bound to start from the source. First, make SiC wafers larger and more refined in order to activate the entire industrial chain.
3.3 SiC substrates: semi-insulating substrates are "precise, expensive and difficult", and there is still room for domestic pursuit
Traditional semiconductor materials such as Si, Ge, GaAs, InP, etc., can be grown using various melt methods, such as common Czochralski method, liquid-sealed Czochralski method, gradient condensation method, Bridgman method, etc. However, the above None of the methods are suitable for the growth of SiC single crystals. Theoretical calculations show that SiC conforming to the stoichiometric ratio can only be melted when the pressure reaches 105 atm and the temperature exceeds 3200 ℃.
The most commonly used method for preparing SiC substrates is Lely's modified method. There are three main technical routes: physical vapor transport method, liquid phase method and high temperature chemical vapor deposition method. The liquid-phase method is currently only in the laboratory stage, and the high-temperature chemical vapor deposition method is expensive, but the future technological progress can continue to be paid attention to.
Source: Huafeng Capital
Physical vapor transport (PVT) is currently the mainstream growth method for 4 or 6-inch SiC single crystals. The density of microtubes can be controlled very low and the resistivity can be controlled. It uses the temperature gradient between the SiC raw material and the seed crystal to sublime the SiC raw material at a high temperature and transport it to the lower temperature seed crystal for crystallization. Because the PVT method is extremely complex, the SiC crystal grown by the PVT method is prone to grow various defects. However, researchers and enterprises in various countries are actively studying, so that the current single crystal size can grow up to 8 inches, and the microtube density can be the lowest. Reducing to zero, the resistivity can be controlled, so that conductive substrates and semi-insulating substrates can be selectively produced.
As far as the preparation process is concerned, the preparation process of the semi-insulating substrate is more difficult, the performance is better, and the price is several times that of the conductive substrate. Intrinsic SiC single crystal exhibits semi-insulating characteristics due to its wide band gap. However, due to the impurities such as N and B introduced during the growth process, as well as the impurities of B and Al contained in the crucible and insulating materials, the resistivity of the SiC single crystal cannot meet the requirements of the semi-insulating type. At present, there are two main solutions for preparing semi-insulating SiC. One is to directly prepare high-purity semi-insulating crystals by strictly controlling SiC raw materials, equipment and growth environment, which requires extremely high level of technology. Another method is to introduce a deep compensation center into the SiC crystal to compensate for the unintentionally doped N, B and other shallow energy level background impurities in the SiC crystal to make the crystal reach a semi-insulating level. In contrast, conductive SiC substrates are relatively easy to prepare. Therefore, the prices of the two products are quite different. According to the reference data, the price of a single piece of 6-inch n-type SiC substrate is 6000-7000 yuan, and the price of a single piece of 6-inch semi-insulating substrate will even exceed 20,000 yuan.
The resistivity and uniformity of the SiC substrate have a great impact on the performance of the device. High-temperature high-frequency, high-power RF microwave devices must use semi-insulating SiC single crystal as the substrate. The main reason is that the buffer layer of devices made of conductive substrates cannot be completely depleted during operation, so the maximum output power is only a fraction of that of semi-insulating substrates. In addition, the thermal conductivity of the high-purity semi-insulating SiC substrate is as high as 4.9 at room temperature, so that the fabricated device has better heat dissipation performance.
After the crystal grows, it is necessary to use diamond abrasives to cut, grind and mechanically polish the SiC crystal, and use a chemical mechanical polishing process for surface treatment. The mechanically polished SiC lens has a certain thickness of processing damage, and the damage layer directly affects the quality of the SiC epitaxial layer, thereby affecting the device performance. The wafer processing technology applied to the first and second generation semiconductors cannot remove the residual damage layer on the surface of the SiC wafer. Therefore, it is necessary to develop a chemical mechanical polishing process to eliminate the residual damage layer on the surface of the SiC wafer.
The problem of overall deformation of large-size SiC wafers needs to be solved with out-of-the-box SiC wafer batch processing technology. Due to the extremely high hardness of SiC, the wafer processing and removal speed is slow, and the processing time is too long, which will cause serious deformation of the SiC wafer, and the warpage and total thickness of the wafer will vary greatly, which is particularly prominent in large-size wafers. The currently developed ready-to-use SiC wafer batch processing technology is used to process the non-damaged layer on the wafer surface, and the wafer Warp and TTV values are well controlled.
International companies still occupy a major position in the SiC substrate production market, 6-inch materials are fully commercialized, and 8-inch materials have entered the pre-production sample preparation stage. Companies with SiC crystal growth, complete equipment research and production scales are Cree, II-VI, DowCorning, SiCrystal, and Rohm. These companies focus on devices made of SiC crystals, and are mainly IDM-type companies. As of 2019, Cree has fully switched to 6-inch SiC products, and the first batch of 8-inch SiC substrates has been sampled, and mass production is expected to be achieved in 2022. SiC crystal growth equipment is also mainly occupied by foreign manufacturers. At present, the manufacturers that can provide crystal growth equipment to the market include AIXTRON, PVATepla, Linn and GTAT of the United States.
Domestically, Shandong Tianyue and Tiankeheda have all realized the industrialization of 4-inch conductive and semi-insulating substrates, but they still have a significant gap with international mainstream companies. Shandong Tianyue is currently expected to mass-produce 6-inch conductive substrates, and the China Electronics Institute 2 has realized the manufacture of 6-inch semi-insulating substrates, and 8-inch is expected to continue to follow up. At present, mainstream domestic manufacturers are capable of preparing low-microtube density substrates. With independent 6-inch N-type 4H-SiC single crystal substrate material technology, the microtube density is reduced to 0.13 pcs/cm2, and 4-inch semi-insulating SiC substrates are commercially available in batches. The surface roughness is 0.082nm, and the supply exceeds 10,000 pieces. In terms of epitaxy, the current commercial size of SiC homogeneous epitaxy is 4-6 inches, the base plane dislocation (BPD) is ≤1cm-2, and the maximum thickness can reach 200μm.
According to Cree's forecast, the SiC substrate market will surge from USD 56 million in 2017 to USD 1.2 billion in 2022, with a CAGR of up to 80%.
Source：Cree
3.4 SiC epitaxy: Cree dominates Small off-angle substrates will become the mainstream
In order to manufacture SiC semiconductor devices, it is necessary to grow one or several layers of SiC film (or GaN) film on the surface of the substrate. These films have different n and p conductivity types. At present, the mainstream method of SiC homoepitaxial growth is chemical vapor deposition.
In the silicon carbide epitaxial growth scheme, the non-off-angle substrate may cause the structure of the grown epitaxial layer to be impure, which greatly affects the performance of the silicon carbide device. Early silicon carbide was grown epitaxially on a non-offset substrate, that is, the angle between the epitaxial surface normal and the crystal axis (c-axis) of the wafer cut from the ingot is θ=0°, such as Si(0001 ) Or C(000) plane, the epitaxial surface has almost no steps, and the epitaxial growth is expected to be controlled by an ideal two-dimensional nucleation growth model. However, the actual growth found that the epitaxy result is far from ideal. Since silicon carbide is a polytype material, it is easy to produce polytype inclusions in the epitaxial layer. For example, the presence of 3C-SiC inclusions in the 4H-SiC epitaxial layer makes the epitaxial layer "impure" and becomes a mixed phase structure. Greatly affect the performance of silicon carbide devices, and even such epitaxial materials cannot be used to prepare devices. In addition, such an epitaxial layer has a large macro-epitaxial defect density, and conventional semiconductor processes cannot be used to prepare devices, that is, the film quality is difficult to reach the wafer-level epitaxial level.
Oblique cutting the substrate with a large deflection angle of 8° is one of the transition solutions to solve the epitaxial problem, and the base plane dislocation BPD is an important parameter to measure its quality. The 8° substrate cutting method means that when cutting the wafer, the epitaxial surface of the substrate is offset by 8° toward the <1120> direction, so that the epitaxial surface forms high-density nano-level epitaxial steps. It was found that a homogeneous epitaxial layer can be prepared at 1500°C, reaching the wafer level. However, this method will introduce the base plane dislocation BPD. The BPD defect density is proportional to the off-angle of the substrate, and the extended Shockley dislocations that it brings will cause unpredictable changes in the performance and reliability of the device.
The epitaxial growth should return to the direction of the substrate with a small deflection angle, and Cree currently mainly promotes the substrate with a 4° deflection angle. Cree has an industry leading position in the preparation of SiC substrates. Its products are the industry's vane and represent the development direction of demand. Now Cree’s main deflection angle is 4° substrate, which has a profound impact.
According to the difference of conductivity type/semi-insulating type and epitaxial material SiC/GaN, the epitaxial wafer with SiC as the substrate can be divided into four types. Corresponding to different device types. As far as SiC homoepitaxial materials are concerned, the largest and first to be used in large-scale applications is the automotive electronics field.
Source: Huafeng Capital
3.5 SiC devices and market: SiCSBD and MOSFET "go hand in hand" to replace silicon-based devices
Compared with silicon materials, SiC materials have the following advantages:
(1) The band gap of SiC is three times that of Si material, which makes the leakage current of SiC devices several orders of magnitude less than that of silicon devices, so the power loss of SiC high-power devices is small;
(2) The thermal conductivity of SiC is about three times that of Si material, which makes SiC devices easier to dissipate heat, thereby reducing the dependence of the circuit system on heat dissipation equipment, reducing the volume of the circuit system, and improving the integration of SiC integrated circuits;
(3) The electron saturation drift speed of SiC is twice that of Si, which makes SiC high-power devices have lower on-resistance compared with Si devices, greatly reducing the conduction loss of power devices;
(4) The critical breakdown electric field of SiC is ten times that of Si, which makes SiC power devices have higher withstand voltage and current resistance, and are more suitable for extreme environments such as high voltage and high power.
According to different structures, SiC power electronic devices are mainly divided into two types: diode (SBD) and transistor (MOSFET).
3.5.1 SiCSBD: Can replace SiSBD in application scenarios below 1200V
The devices made by SiC homoepitaxial can obtain high withstand voltage diodes above 600V with the SBD (Schottky barrier diode in Chinese) structure of the high-frequency device structure. In contrast, the highest withstand voltage of mainstream SiSBD products is only It is 200V.
Source：Rohm
SiCSBD can be widely used in power factor correction circuits (PFC circuits) and rectifier bridge circuits in air conditioners, power supplies, power regulators in photovoltaic power generation systems, and fast chargers for electric vehicles. If SiC-SBD is used to replace the current mainstream fast PN junction diode (FRD: Fast Recovery Diode), the recovery loss can be significantly reduced. It is conducive to the high efficiency of the power supply, and the miniaturization of passive components such as inductors through high-frequency driving, and noise reduction.
Source:CASA research
The preparation principle of SiC SBD is to use n+ type 4H-SiC substrate (the mainstream product is 350μm thick), and the carrier concentration is 5E18/cm3. Then a layer of n-type conductive 4H-SiC epitaxial film is grown on it, the thickness of which is generally 6 μm or 12 μm, and the carrier concentration is 1015-1016/cm3.
It is difficult to prepare SiC SBD with a withstand voltage higher than 1000V, and it is necessary to switch to a MOSFET structure. To prepare a SiC SBD with a single tube withstand voltage higher than 10,000V, it is necessary to grow an epitaxial layer with a thickness of 120-200μm. At present, the growth rate of commercial silicon carbide epitaxial equipment is generally 10 μm per hour, and it cannot grow continuously. Generally, it can only grow for 3-5 hours. Therefore, it is very difficult to grow materials with a thickness of 100 μm or more, which limits high-pressure applications. In addition, if the high-power device still has a thick conductive layer substrate, its operating resistance will not be ignored.
The current commercial SiCSBD highest withstand voltage level has reached 3300V, but more than 90% of the product withstand voltage range is still concentrated in 650V and 1200V, the working current is concentrated below 60A, while 1700V and 3300V products are very few. As commercial SiCSBD products currently cover most of the application requirements, the number of new products launched in 2019 has decreased. It is worth mentioning that the 4 SiCSBDs launched in 2019 are all compliant with the AEC-Q101 standard, which can be applied to the power electronic device market in the fields of new energy vehicles and photovoltaics.
3.5.2 SiCMOSFET: High-voltage application scenarios replace SiMOSFET
The preparation of SiC MOSFET is different from Si. The difference between Si and 4H-SiC is that the former can use bulk materials as the pressure-bearing layer to prepare MOSFETs, while the latter cannot. It is necessary to prepare similar bulk materials HPSI or n (n channel-channel) or p by epitaxial coating by CVD technology. -(p-channel) as a pressure-bearing layer, and the substrate used needs to be thinned.
Source: CNKI
As a voltage-controlled unipolar device, power MOSFET has the advantages of small input current, high power gain, high switching frequency, and its current has a negative temperature coefficient, so it is easy to operate in parallel, without secondary breakdown, and has a wide range Safe working area, so it is welcomed by industry and commerce.
Source：CreeWolfspeed
The core performance indicators of the MOSFET include the maximum allowable voltage VDS between the drain and the source, the drain current ID, and the power consumption Ptot. Among them, power consumption is the core indicator to measure device performance.
MOSFET power consumption comes from static conduction losses and dynamic switching losses. The conduction loss is affected by the on-resistance RDS(ON) between the drain and the source. When the package conditions are the same, the higher the VDS, the larger the corresponding on-resistance RDS(ON), and the greater the conduction loss. The switching loss is affected by parameters such as gate charge and parasitic capacitance, and the gate charge Qg is inversely proportional to the on-resistance RDS(ON).
The core of the MOSFET design upgrade is to reduce the on-resistance as much as possible, thereby reducing the size and reducing the conduction loss; at the same time, it seeks to balance the on-resistance and gate charge.
Source：Rohm
Compared with SiMOSFET, SiCMOSFET is more withstand voltage, and has lower conduction loss and switching loss. Because the SiC breakdown field strength is nearly 10 times that of Si, the MOS device has a higher withstand voltage, a thinner semiconductor layer, and the resistance of the drift zone is reduced to 1/300, which ultimately reduces the conduction loss and reduces the power consumption and size of the device. . Si semiconductor devices of 1200V and above mostly adopt IGBT structure, but the switching frequency of IGBT is low due to the influence of tail current.
SiCMOSFET will become the main driving force for the rapid expansion of the SiC market in the next five years. In 2019, the highest withstand voltage of commercial SiC MOSFET is 1700V, the working current is below 65A, and there are mainly four voltage levels of 650V, 900V, 1200V and 1700V. According to Yole's prediction, SiCMOSFET will become the main driving force for market expansion in the next five years.
Source：CASA research
1700SiC module is expected to replace SiIGBT module in performance. In 2019, Mitsubishi newly launched a full SiC module with a maximum operating voltage of 3300V; Rohm developed a 1700V/250A full SiC module, whose pressure standard is already higher than the current demand for new energy vehicles. It is expected to be used in outdoor power generation systems and industrial high voltage Play a role in the power supply.
At present, the main players in the field of SiC devices are still international semiconductor giants such as ON Semiconductor, Roma, Infineon, STMicroelectronics and Cree. According to Mouser data, the SiCSBD products produced by these five companies account for nearly 70% of the global market.
Domestically, there is basically no difference in the voltage levels covered by domestic and foreign products for SiC devices. SiC SBD covers the voltage range of 600V-3300V. At present, Tyco Tianrun has achieved mass production of 600~1200V/1~50ASiCSBD, and has both 1700V and 3300V voltage grade products. The commercial SiC SBD currently has a maximum withstand voltage of 6.5kV and a working current of 25A, realizing mass supply; in terms of SiC MOSFET, a variety of 1200V high-power devices and module products applied to the motor drive of new energy vehicles have been launched, and the full SiC power module has the highest The specification is 1200V/600A. Important breakthroughs have been made in technologies such as automotive SiC motor controllers, automotive independent 1200V SiC chips and modules, and automotive high-temperature and high-current SiC MOSFET double-sided silver chips. At the same time, a 60kW full SiC DC charger has been developed with an overall efficiency of 96%, which is 2% higher than the silicon-based prototype; a 400kW full SiC DC charger prototype that meets the application of bus charging stations has been produced; and the domestic SiC charging equipment has been realized. Mass production.
3.6 SiC device application: automotive electronics is the biggest breakthrough
The trend of new energy vehicles is positive, and the value of power semiconductors is prominent. The components involved in the application of power semiconductors in the new energy vehicle system architecture include: motor drivers, on-board chargers (OBC)/off-board charging piles, and power conversion systems (on-board DC/DC). With the development of new energy vehicles, the demand for power semiconductor devices will increase day by day. According to Infeneon's statistics, the average value of semiconductor devices used in a traditional fuel vehicle is US$355, while the value of semiconductor devices used in new energy vehicles is US$695, almost doubled. The increase in power devices is the most significant, from US$17 to 265. The dollar has increased nearly 15 times.
The introduction of SiC MOSFETs in automotive electronics has significantly increased the power density of automotive power systems, reduced switching losses, and improved the thermal management of electric vehicle OBC.
Source：Cree
According to Cree's calculations, from the cost of the system bill of materials, the use of SiCMOSFETs can reduce the cost of electronic components of the vehicle by 13% compared to Si-based devices.
Source：Cree
The main application areas of SiC in automotive electronics are shown in the figure above. The main application scenarios are:
(1) Off-board DC fast charger
The external charger can convert the incoming external alternating current (AC) into the direct current (DC) power supply mode required by the EV ecosystem and store it in the battery.
The application of SiC power devices in the field of charging piles can improve the switching frequency and efficiency of the power system, and reduce the weight and volume of passive devices (inductors, capacitors, etc.), and improve system efficiency. Taking the three-phase power conversion system of the electric vehicle charging pile facility as an example, a 1200V SiCMOSFET can construct an LLC full-bridge stage for the DC-DC conversion stage. The Si solution relies on 650V Si super junction MOSFETs, which usually require two series LLC full bridges to support the 800V DC link, and four sets of SiC MOSFETs plus driver ICs can replace eight sets of Si super junction MOSFETs plus Driver IC. In addition, the SiCMOSFET can greatly improve the efficiency, and only two switching positions (four for Si bases) need to be opened in each conduction state, thereby improving the efficiency of the charging cycle.
STMicroelectronics is developing a charging board solution with a maximum of 40KW. The charging station can assemble multiple power boards to make the output power of the charging pile reach more than 200KW, so that the vehicle can be fully charged in less than an hour. In the future, if the output power can reach more than 350KW, the time to fully charge an electric vehicle with a power of 90KW is expected to be shortened to tens of minutes.
(2) Car battery charger OBC
This component converts the DC power of the battery subsystem into the AC power of the main drive motor. When the vehicle receives external power from the grid, the rectifier circuit of the device converts AC power to DC power to charge the battery. The system also collects the kinetic energy generated by the vehicle's momentum through regenerative braking and sends it to the battery.
The introduction of silicon carbide can make each stage of the charging process faster, while reducing energy consumption and heat. According to Cree's calculations, the energy loss and heat generation of SiC car chargers are reduced by 60% compared to Si car chargers.
Another example comes from the OBC system design of the EV battery system. OBC design for EV battery systems is another use case. 650V SiC MOSFETs provide a competitive advantage by enabling designers to increase efficiency while enhancing the ability to support bidirectional power flow without reducing weight, size, and design complexity. This greatly reduces the size and weight of the OBCs, which convert AC power from the grid to DC power for the batteries in the vehicle.
The SiC-led two-way OBC system allows electric vehicles to charge other electric vehicles and other AC equipment. Unlike the one-way OBC that is idle in the vehicle when it loses charge, the two-way OBC not only pulls the charge from the grid, but also replenishes the charge. This bidirectionality also allows the end user to provide energy to other AC power equipment, or to provide another vehicle that has exhausted its battery with an EV equivalent to "jump start".
The market prospects for SiC car chargers are broad, and major auto manufacturers have deployed one after another. According to Yole statistics, as of the end of 2018, more than 20 car manufacturers around the world have adopted SiC SBD or SiCMOSFET devices in their car chargers. Among them, the industry pioneers are BYD and Tesla. Geely, Volkswagen, Renault, Nissan and others have also deployed. It is expected that until 2023, the market will continue to maintain a growth rate of 44%.
Source：Yole
(1) EV powertrain/inverter, power control unit
For EV powertrains, high-power SiCMOSFETs can reduce leakage power compared to Si-based devices, which reduces power loss under high current conditions. Compared with IGBT inverters, SiC MOSFET inverters can reduce switching losses by 80%, and the inverters are directly integrated inverters without the need to install additional liquid coolers. At the same time, the thermal management performance of the SiC MOSFET inverter is better. According to Cree’s calculations, SiC inverters can increase battery life by 5-10% and save US$400-800 in battery costs (80kWh battery, US$102/kWh). After offsetting the cost of new SiC devices of US$200, they can achieve at least A $200 bicycle cost reduction. Each inverter in Model 3 contains 48 power SiC MOSFETs, which reduces the body of Model 3 by 20% compared to Model S.
For the power control unit PCU, SiC can reduce the power loss of silicon-based semiconductor devices when high voltage and strong current are applied. Replacing the Si diodes equipped with traditional PCUs with SiC diodes and Si IGBTs with SiC MOSFETs can reduce the total energy loss by 10%. At the same time, the device size can be greatly reduced, making the vehicle more compact.
According to Cree's forecast, the market for SiC in automotive electronics will exceed US$2.4 billion in 2022, a nearly 350-fold increase from the US$7 million market space in 2017.
Figure: SiC market in the global automotive electronics industry Source：Cree
Looking at 2019, more than 20 automobile manufacturers in the world have used SiC devices in on-board chargers (OBC), and Tesla Model 3 inverters use full SiC power modules produced by STMicroelectronics. In terms of vehicle charging infrastructure, Delta has teamed up with GM to develop a 400kW ultra-fast charging system (XFC) that uses SiC power semiconductor devices. In terms of electric drive, Cree also teamed up with ZF to promote cooperation in the field of electric drives and promote the development of electric drive powertrains using SiC-based inverters.
3.7 Total SiC device market: 2020-2022 will welcome rapid growth
SiC solutions for automotive inverters will become the main driving force for the growth of the SiC device market. According to Yole's forecast, the global SiC device market will maintain a compound annual growth rate of as high as 40% from 2020 to 2022, and will exceed the threshold of $1 billion. Among them, SiC for automotive inverters will benefit from the preferential policies of Asian and European governments for electric vehicles. According to ROSKILL&UBS forecasts, it is estimated that within ten years, the market share of electric vehicles will increase from less than 2% to 15%, which will inevitably expand the demand for SiC charging equipment for automobiles.
With the construction and improvement of SiC's high-performance passive components, drivers, and sensor equipment and ecosystems, SiC will cover the traditional Si market. With the opening of the traditional Si-based market to SiC, it is bound to intensify the competition in the SiC industry chain from the substrate, epitaxy, and downstream devices to the application side. It is estimated that 10% of the Si-based device market will be covered by the end of 2025. In the longer term, SiC's 3300V-10KV products are expected to be deployed in MW motor drives, high-voltage DC power conversion, solid-state circuit breakers and other applications for railway traction, wind, marine and industrial applications.
3.8 Packaging: Discrete devices still use traditional packaging. The high-frequency era will be replaced by co-packaging technology
Most SiC discrete devices still use traditional Si-based packaging technology. However, at the beginning, SiC discrete devices still used traditional silicon-based packaging technology, such as TO247-3L. With the introduction of 650V SiC switching devices, more Si packaging technologies are expected to be adopted, such as DFN8x8, TOLL, DPAK-3L and D2PAK-3L.
Figure: SiC discrete devices have widely adopted discrete surface packaging and full-hole packaging technology
Source：UnitedSiC
The research and development of SiC full module packaging technology has made considerable progress, and the future can be expected. At present, companies such as Mitsubishi and Fuji can already provide SiC module packaging technology based on IPM technology. Automotive electronic-grade IPM packaging technology with built-in gate drivers can minimize high-frequency switching losses. Although the cost is still high at present, it can provide good performance for 8-25KW power supplies and on-board chargers.
3.9 Outlook
The entire SiC industry chain is constantly adapting, gradually approaching the critical point of the breakthrough development of the SiC industry. At the upstream end, the 6-inch substrate material is mature, and manufacturers have a relatively in-depth understanding of the material characteristics and defect control of SiC; in terms of design and packaging, the ecosystem of excellent gate drivers as the core simplifies device design and improves Modular packaging technologies such as discrete packaging and IPM are also constantly improving; from the perspective of the entire industry, preparations for large-scale applications are also steadily advancing.
All-round progress makes SiC devices have much more room for imagination than Si-based devices. Among them, the power supply and car charger fields are expected to become the first development area for high-speed penetration of SiC.
Overall, the development speed of the SiC industry will benefit from the first breakthrough in upstream substrate material technology and the optimization of device design and packaging. The development is "one step faster" than GaN.
4. In recent years, the third-generation semiconductor industry policy and investment situation
The following sources of information: Comprehensive public information and the 2019 Third Generation Semiconductor Industry Report
4.1 Policy
Source: Huafeng Capital
Source：IEEE
