GaN new breakthrough: faster car charger, mass production expected?

Today, R&D technology for power semiconductors based on new materials such as silicon carbide (SiC) and gallium nitride (GaN) is attracting attention. Professor Yusuke Mori of Osaka University has been working on semiconductor R&D based on the Japanese Ministry of the Environment's policy of "accelerating the application and popularization of materials (gallium nitride) and CNF (carbon nanofibers) to further achieve carbon neutrality.

Today, power semiconductors using wide bandgap materials have begun to be put into practical use. It is reported that Tesla's motor (Inverter) in the United States (Inverter) uses silicon carbide semiconductors. In addition, many readers should have seen some extremely small AC converters (AC converters) using gallium nitride semiconductors in home appliance sales centers. In high-voltage operation, the electrical performance and effectiveness of the internal circuits of power semiconductors made of wide bandgap materials are much higher than those of conventional semiconductors made of silicon.

For silicon carbide semiconductors and gallium nitride semiconductors that have been put into practical use, the application terminal has different requirements for their rated voltage (higher than the rated voltage, which is the basic voltage to maintain reliability), respectively, as follows, silicon carbide withstand voltage above 1000V, gallium nitride withstand voltage below 1000V. Based on the above distinction, a "wordless tacit understanding" has been formed between power semiconductor manufacturers and R&D companies.

However, the above situation is likely to change. Because gallium nitride materials can greatly reduce the defect (misalignment) density of wafers, so they can improve the performance and efficiency of application terminals, and are far superior to silicon carbide materials, GaN is expected to achieve large-scale mass production. Today, developers are working to accumulate data to confirm these conclusions. Professor Yusuke Mori of Osaka University in Japan is at the forefront of these R&D activities.

Although gallium nitride power semiconductors are extremely applicable, they still face three social problems

From the perspective of physical properties alone, gallium nitride is more suitable for power semiconductor materials than silicon carbide.

The researchers also compared the "Baliga performance index (the performance value of semiconductor materials relative to silicon, that is, silicon is 1)" of silicon carbide and gallium nitride, which is 500 for 4H-SiC and 900 for gallium nitride, which is extremely efficient. IN ADDITION, SILICON CARBIDE HAS AN INSULATING FAILURE ELECTRIC FIELD STRENGTH (INDICATING THE VOLTAGE RESISTANCE CHARACTERISTICS OF THE MATERIAL) OF 2.8MV/CM, AND GALLIUM NITRIDE IS HIGHER, 3.3 MV/cm. In general, the power loss in low-frequency operation is the third power of the insulation damage electric field, and the power loss in high-frequency operation is the 2nd power of the insulation damage electric field, which is inversely proportional, so the power loss of gallium nitride is lower (higher work efficiency).

So, why is the practical application of silicon carbide earlier than gallium nitride in high-voltage applications? The reason is as follows: Silicon carbide is easier to form silicon dioxide (SiO2) when making MOS FETs, and "GaN wafers face three major problems" (Professor Mori). (Figure 1 below)

Figure 1: Problems faced by GaN wafers listed by Yusuke Mori of Osaka University, Japan. (Image courtesy of Osaka University, Japan)

The first problem is that due to the small size of the Bulk Wafer, only low-cost wafer products can be produced before, and some products cannot even meet the test requirements. Traditionally, only 2-inch wafers can be produced, and now it is finally possible to produce 4-inch wafers. The industry generally believes that only large-size wafers of more than 6 inches can meet the mass production needs of power semiconductors, so it has not yet reached the requirements for mass production. In addition, the GaN power semiconductors used in the small AC converters mentioned above are made on a chip that forms a GaN layer on a silicon (Si) substrate with a maximum size of 6 inches. However, due to the difference in the crystallization constant (Lattice Constant) of silicon and gallium nitride, the defect density of the gallium nitride layer is high, and it is impossible to form a longitudinal FET that can withstand high voltage and high current, nor can it produce a high-performance horizontal HEMT.

The second problem is that the GaN wafer itself as a bulk is not of high quality. Today's crystalline wafers have a maximum dislocation density of up to 106/square centimeter, a level of density that is not suitable for power semiconductor production. However, the tilt angle (Off) distribution of a 2-inch wafer (which is an indicator of wafer warpage) is 0.2 degrees, making it difficult to achieve large size and low cost. However, the low-quality wafers described above are suitable for the production of optical semiconductors. However, for power semiconductors, current needs to circulate in most of the wafer, so misalignment defects are the main reasons for high voltage resistance, current flow, and low production yield. To be suitable for power semiconductors, the following misalignment density requirements need to be met: the high voltage range needs to be 0.65~3.3kV, the current of a single chip (Chip) is more than 100A, and the production yield must reach 90% (low misalignment defects and low warpage must be achieved).

The third problem is that wafers are expensive. At present, the price of a 2-inch wafer is 100,000 yen to 200,000 yen (about 5,220 yuan to 10,440 yuan). Here's why it's so expensive: there hasn't been a technology in place to produce large-size wafers with high yields. Only wafers with a size of 6 inches and a price of less than 100,000 yen (about 5,220 yuan) are suitable for mass production of power semiconductors.

Successful acquisition of high-quality, large-size GaN wafers suitable for mass production of power semiconductors

The root cause of the above problems faced by gallium nitride wafers lies in the growth method of gallium nitride crystals. Today's mass-produced bulk gallium nitride wafers are produced as follows, and gallium nitride crystals are generated by a gas-phase epitaxy method called HEPV (Hydride Vapor Phase Epitaxial, hereinafter referred to as: "HVPE") on a sapphire substrate. If sapphire, etc. is used as the base material for crystal growth, a large number of misalignment defects will occur because the nitrogen sapphire material is different from the Lattice Constant of gallium nitride. In addition, with "HVPE", crystals are generated at a high temperature of 1000 degrees, so when cooled at room temperature after growth, the entire wafer will warp and the tilt angle (Off) will appear.

In addition, there is a crystallization method called "Ammono-thermal" that produces high-quality crystallization, unlike the "HVPE" method used in the mass production process of bulk gallium nitride wafers. As a method for generating artificial crystal crystals, the "ammonia thermal method" adopts the hydrothermal synthesis method (which has achieved industrial application). Increase the temperature and pressure of ammonia in the pressure vessel to make it in a supercritical state, dissolve gallium nitride polycrystalline, and then precipitate single crystals on gallium nitride seed crystals. Using gallium nitride seed as the base material and using liquid phase growth method, high-quality single crystals can be produced. "However, with the ammonia heat method, growth stops as soon as a stable surface appears during the crystallization growth process. Based on the above phenomenon, although 4-inch wafers can be made, it takes time to produce larger wafers. (Professor Mori)

However, it was not possible to produce high-quality GaN wafers in the past, but the situation has improved significantly in recent years. Technologies have been established to manufacture high-quality, low-cost GaN wafers. Osaka University and Toyoda Gosei Co., Ltd. have jointly developed a new technology that solves these problems (Figure 2 below), which combines the Na Flux method (sodium co-solvent method, which uses this method to grow gallium nitride crystals)" and the "Point Seed method, which uses this method to achieve large-size wafers)".

Figure 2: The fusion of the "Na Flux method" and the "Point Seed method" made it possible to produce large-size bulk GaN wafers. (Image courtesy of Osaka University, Japan)

"Na flux (sodium solubilization) method" refers to the exposure of sodium/gallium solution to nitrogen with a gas pressure of 30 to 40, dissolving nitrogen in the solution and making it saturated, so that gallium nitride crystallization precipitates. This is a technology developed in 1996 by Professor Hisanori Yamane of Tohoku University. The "Na Flux (Sodium-assisted Solitude) Method" is characterized by the fact that high-quality crystals can be formed on the surface of the seed crystals even if the crystal quality is low. However, this method alone, although it is possible to form perfect crystals by relying on a small dot, it is not possible to form large-scale crystals. Therefore, the "Point Seed Method" is used to form large-sized wafers. That is, the crystal seeds are distributed in a large area on the large substrate, and in the process of crystal growth, they are combined separately to form a single crystal.

According to Professor Sen, the above method can be used to obtain ideal crystallization suitable for mass production of power semiconductors, with a misalignment density of 104/cm² or less and a 6-inch wafer tilt angle distribution of 0.2 degrees. In addition, a 6-inch Bulk GaN substrate (the largest in the world) has been successfully made. Moreover, if a larger size substrate and more seeds are used, 10-inch wafers can be produced without reducing the throughput.

In addition, there is another method, that is, using a bulk-type GaN substrate as the seed crystal, using the "ammonia thermal method" to produce a high-quality, large-size bulk-type substrate (Figure 3 below). In response to the above method, Professor Sen pointed out: "The cost is comparable to existing silicon carbide substrates, and larger sizes can be achieved. Companies such as Osaka University and Toyoda Gosei Co., Ltd. have participated in the Reiwa 4-year project "Reiwa 4 to further achieve carbon neutrality and accelerate the application and popularization of components and materials" proposed by the Ministry of the Environment, and Mitsubishi Chemical Co., Ltd. (with "ammonia thermal method" technology) has recently joined the project, and the participation of many companies will be more conducive to the implementation and verification of the project.

Figure 3: Fusion of "Na Flux (sodium solubilization) method" and "ammonia heat method". The advantage of the "Na Flux method" is that the wafer can achieve a large size and higher quality; The advantage of the "ammonia thermal method" is that it improves wafer quality. After the fusion of the two, gallium nitride wafers can be obtained at a lower cost than silicon carbide. (Image courtesy of Osaka University, Japan)

It can successfully improve the performance and yield of components

According to Professor Sen, the performance and yield of gallium nitride elements were generally improved by using gallium nitride substrates made of "Na flux method" and "point seed crystal" method.

Osaka University and Panasonic Group collaborated to fabricate longitudinal gallium nitride FETs based on bulk substrates using the Na flux (sodium solubilization) method, and examined the yield from the perspective of chip OFF performance (Figure 4 below). The yield of chips made from commercially available gallium nitride substrates is only 33%, and the yield can be greatly increased to 72% by using the above method. In addition, the above results are obtained based on laboratory foundations, and there is still a lot of room for improvement in the future.

Figure 4: High-quality, large-size silicon nitride substrates can be produced using the "Na Flux method" and the "Point Seed Method". (Image courtesy of Osaka University, Japan)

In addition, researchers have begun to use the "OVPE method (Oxide Vapor Phase Epitaxy, abbreviated: OVPE, which can be used to make ultra-low resistance wafers, developed by Osaka University in Japan, and promoted by the Panasonic Group to promote its practical application)" to grow gallium nitride crystals on seeds made of "Na Flux method" and "Point Seed Crystal" to develop higher performance longitudinal GaN FETs. The resistance of the finished wafer is about 10-4 Ωcm², which is much lower than that of silicon carbide wafers (about 10-3 Ωcm²), the misalignment density is 104/cm², and the GaN film thickness is more than 1 mm. The researchers obtained a wafer that promises longitudinal FET. Compared to silicon carbide-based longitudinal MOS FETs, longitudinal FETs have higher potential in terms of performance (Figure 5 below). Compared to chips made of conventional bulk GaN wafers, the ON resistance value of the experimentally fabricated diode is reduced by 50%, and the OFF resistance value of the longitudinal FET is reduced by 15% (or more).

Figure 5: Performance vs. wafer characteristics of power semiconductors. By using the "OVPE method", the resistance of the wafer can be reduced. (Image courtesy of Osaka University, Japan)

In the project of the Ministry of the Environment, Osaka University in Japan is focusing on the development of GaN substrates with ultra-low resistance, high quality, large size and related products and modules in order to realize the application in electric vehicle drive inverters. (Figure 6 below)

Figure 6: Development plans for ultra-low resistance, high-quality, large-size GaN wafers, and related applications and modules. (Image courtesy of Osaka University, Japan)