Technical Insight
Silicon carbide material, devices and applications: evaluating the current market and future trends
There are strong opportunities for growth in the SiC market, particularly in electronic applications, provided that material quality continues to improve. Philippe Roussel of Yole Developpement reports.
While silicon carbide material was initially used for its mechanical properties, the production of SiC ingots, developed in the 1990s, opened up new applications in the fields of power electronics and optoelectronics.
SiC is used to manufacture devices in the power electronics and RF/microwave field, as well as being used as the substrate for epitaxial growth of nitride-based device structures. In optoelectronics, SiC is again used as the substrate material for the heteroepitaxy of nitride layers for LEDs and lasers. Finally, in the MEMS and microsystems fields, it is considered as a suitable material with which to design microsystems dedicated to harsh environments, because of its thermal and chemical properties.
Bulk crystal growth of SiCAs summarized in table 1, numerous techniques have been developed to grow SiC ingots. These are as follows:
Sublimation In the most commonly used technique, ultra-pure SiC powder is sublimed from a crucible in order to deposit SiC material along the length of a crystalline SiC seed. This method is used today to produce 4H and 6H SiC crystals.
High-temperature chemical vapor deposition (HT-CVD) This process is carried out at about 2200 ºC and involves the recombination of two gases, for example silane and propane, to form a thick SiC layer and then a SiC ingot. This method is just reaching the production level and is mainly used to manufacture 4H semi-insulating SiC wafers.
Heteroepitaxy of SiC on silicon In this procedure, the silicon is used as a sacrificial substrate, allowing epitaxial growth of cubic SiC. Once the SiC layer is thick enough, the silicon substrate is removed. Sample wafers are now available, and large-diameter substrates are expected to be obtained by this technique.
Vapor liquid solidification (VLS) This R&D-level technique is based on the "conversion" of a silicon surface in the liquid phase under a carbon atmosphere to produce SiC at the interface. The process begins with carbonization of the silicon surface (using a standard SiC CVD process at ~1500 ºC). The sample is then placed in a carbon atmosphere to start the conversion process. Due to the cubic silicon substrate, the SiC is also cubic (3C).
Other new techniques are also under development. For example, it was recently proved possible to combine sublimation with HT-CVD to achieve high-quality material with a high growth rate. Also, the Smart Cut technology developed by French company Soitec makes it possible to transfer a thin SiC layer from a high-quality wafer to a low-quality SiC substrate. The final material is a bulk conductive or semi-insulating SiC wafer.
Typical SiC growth rates range from 100 µm/h to values in the order of mm/h, depending on the technique. Sublimation or HT-CVD processes provide ingots of about 20 mm in length, and each ingot is expected to give 5-20 SiC wafers depending on the polytype and orientation.
Market issues for SiC wafersAlthough larger wafers have been demonstrated, SiC material is currently only commercially available in diameters of up to 3 inches. Moreover, it suffers from defects that limit the range of applications for components, as shown in table 2.
Micropipes are larger defects that look like cracks in the bulk material and limit the available active area of the wafer. High-power-density components require large micropipe-free areas, but current micropipe densities only allow active areas of about 20 mm2.
According to numerous component manufacturers, the compatibility of 4 inch wafers with standard processing equipment makes the realization of this wafer diameter a key issue if the SiC market is to grow substantially.
With an estimated 250,000 SiC wafers (mainly 2 and 3 inch) produced in 2003, the SiC wafer market is at the start of a steep growth curve. We estimate that the volume of SiC wafers will reach more than 600,000 units in 2007.
At present, SiC wafer production is dominated by Cree, which has about 85% of the market. However, challengers such as Okmetic, Hoya and Soitec are using new growth techniques, and are expected to make inroads. Due to Cree, Dow Corning (formerly Sterling), II-VI Inc (formerly Litton Airtron) and several smaller companies, around 94% of SiC wafer production is carried out in the US, with 4% in Asia and 2% in Europe.
Until now, 90-95% of SiC wafers have been consumed in optoelectronics applications, where they are used as substrates for blue and green LEDs and violet laser diodes (figure 1). However, other substrates, in particular sapphire and also silicon or even free-standing or bulk GaN, have been used to grow nitride-based epitaxial structures.
Nitride devices that use SiC as a substrate (as fabricated by Cree and Osram Opto Semiconductors) represent about 25% of the total nitride optoelectronics market. Since no new players are entering the market using SiC substrates, we expect the SiC market share in optoelectronic applications to remain the same or decrease.
Power electronicsThe challenge for SiC is to move from the optoelectronics market to power electronics, where components can benefit from the material properties of SiC, resulting in high power densities and high-temperature operation.
Depending on the power range, many SiC components can be manufactured using semiconducting or semi-insulating material (mainly 4H or 3C polytypes; figure 2). The first commercial SiC components were Schottky diodes and MESFET RF transistors.
SiC Schottky diodes are the most advanced components using SiC heterostructures as the active layers. These components are currently in production at an estimated level of 5 million units in 2002. This is a very low figure compared with the production of standard silicon Schottky diodes; one single European producer, STM, manufactured 220 million silicon Schottky diodes in 2002.
The only two producers of SiC Schottky diodes are Infineon (through its SiCed R&D lab) and Cree, in collaboration with Microsemi. A number of other companies including Fuji, Hitachi, International Rectifier, Matsushita, STM and Toshiba have strong R&D programs related to SiC Schottky diodes, but have not commenced production.
Schottky diodes target the high-speed, high-power-density switching market. This includes products or functions like high-frequency power supplies, power factor correction and power conversion in motor controls or power management appliances. The use of SiC components will dramatically reduce the number and the size of all passive components in such systems.
Today, SiC Schottky diodes are mainly targeting the market for components operating at 6-8 A and 600 V, for power factor correction in high-end AC/DC power supplies and uninterruptible power supplies (UPSs). The total accessible market for this segment is estimated to be 200 million parts per year.
The next promising market for SiC diodes is low-end power supplies for use with PCs and home appliances. These markets will be accessible only when the average selling price (ASP) of the SiC parts falls below $1.
Technical and financial bottlenecks are still slowing down the emergence of commercial products. The SiC material accounts for about 40% of the cost of a typical SiC Schottky diode, a further 50% is processing costs, and the remaining 10% is packaging and test. Due to the cost of materials and the need for highly specialized equipment and processes, components manufacturers are hesitating to launch large-scale development in this field.
On the other hand, potential end users such as manufacturers of power supplies or UPS systems, communication base station providers, and rail traction and automotive manufacturers are waiting for new and improved solutions to manage high-power-density systems. However, implementation of SiC components is not a chip-to-chip replacement and requires a redesign of the electronic circuits.
In the RF/microwave field, SiC MESFETs are competing with GaN HEMTs to replace incumbent silicon LDMOS technology in base stations for mobile phones and broadcast (UHF band) communications. High-frequency components are also expected to be used in the L and S bands for military applications such as radar systems and jamming. Microwave heating could become a huge market if the ASPs of SiC MESFETs become compatible with the cost of the currently used magnetron technology.
SiC MESFETs have entered a pre-production level at Cree, Thales and others, but lifetime and production yield remain key issues when moving to full production.
Conclusions and future trendsSiC material offers a wide range of characteristics, with three polytypes covering the main applications. Three inch semiconducting SiC and 2 inch semi-insulating wafers are now available, but the real breakthrough will be the emergence of 4 inch wafers that are compatible with standard semiconductor tool sets.
There are three production techniques for growing SiC (sublimation, HT-CVD and heteroepitaxy), but micropipe density remains too important to manufacture numerous large devices (>2 kV) on a single wafer.
The SiC material market certainly suffers due to the Cree monopoly, with the company acting as a reference for quality and price level. No clear challengers are emerging, but the market will benefit from having a robust second source.
In terms of markets, optoelectronic applications are the leading consumer of SiC material, and consumption is forecast to increase by more than 20% per year. On the other hand, blue LEDs are now available on silicon (although brightness and yield must be improved), the first violet laser diodes on the market use sapphire substrates, and major R&D efforts are focusing on the development of free-standing GaN. This situation means that SiC will lose market share in the future, even if it shows the best lattice mismatch with nitrides and allows the growth of devices with back-face contacts. The market for optoelectronic devices grown on SiC is forecast to reach $500 million in 2003 and $1.2 billion in 2007, a compound annual growth rate (CAGR) of 24% (figure 3).
The alternative market is power electronics, where SiC enables a high switching capacity at high frequency, with no recovery time, compared with silicon. Only Schottky diodes have reached commercial production to date. The cost of these components allows SiC Schottky diodes to target high-end markets with volumes of about 200 million units per year. By reducing ASPs, Schottky diodes will be able to target the PC power-supply market and low-end, cost-driven applications. We consider that the 2003 market for power devices is worth about $60 million, and will grow to around $500 million in 2007, at a CAGR of 70%.
The market is waiting for full SiC switching cells (SiC diodes and SiC transistors) for applications such as 42 V automotive systems, but SiC MOSFETs are not yet ready, and only hybrid solutions (SiC diodes and silicon transistors) are emerging.
For high-power (>2 kV) devices, pin diodes and BJT or GTO devices are not compatible with high micropipe-density wafers and the quality of thick (>100 µm) SiC epitaxial layers has to be improved.
In the RF field, a single SiC (or GaN) transistor is supposed to replace 10 standard transistors (mainly silicon LDMOS or GaAs transistors) as used in today s base stations. This market represents a potential for 1 million components per year including defense applications. The next challenge will be the replacement of magnetrons in microwave ovens. The market for RF transistors using SiC or GaN is expected to reach $400 million by 2007.
However, SiC components do not offer chip-for-chip replacement, and systems have to be redesigned in order to accept SiC components. This constitutes the main challenge for SiC component manufacturers, who will have to teach their own customers (the end users) how to implement this new component in their systems.
SiC is used to manufacture devices in the power electronics and RF/microwave field, as well as being used as the substrate for epitaxial growth of nitride-based device structures. In optoelectronics, SiC is again used as the substrate material for the heteroepitaxy of nitride layers for LEDs and lasers. Finally, in the MEMS and microsystems fields, it is considered as a suitable material with which to design microsystems dedicated to harsh environments, because of its thermal and chemical properties.
Bulk crystal growth of SiCAs summarized in table 1, numerous techniques have been developed to grow SiC ingots. These are as follows:
Sublimation In the most commonly used technique, ultra-pure SiC powder is sublimed from a crucible in order to deposit SiC material along the length of a crystalline SiC seed. This method is used today to produce 4H and 6H SiC crystals.
High-temperature chemical vapor deposition (HT-CVD) This process is carried out at about 2200 ºC and involves the recombination of two gases, for example silane and propane, to form a thick SiC layer and then a SiC ingot. This method is just reaching the production level and is mainly used to manufacture 4H semi-insulating SiC wafers.
Heteroepitaxy of SiC on silicon In this procedure, the silicon is used as a sacrificial substrate, allowing epitaxial growth of cubic SiC. Once the SiC layer is thick enough, the silicon substrate is removed. Sample wafers are now available, and large-diameter substrates are expected to be obtained by this technique.
Vapor liquid solidification (VLS) This R&D-level technique is based on the "conversion" of a silicon surface in the liquid phase under a carbon atmosphere to produce SiC at the interface. The process begins with carbonization of the silicon surface (using a standard SiC CVD process at ~1500 ºC). The sample is then placed in a carbon atmosphere to start the conversion process. Due to the cubic silicon substrate, the SiC is also cubic (3C).
Other new techniques are also under development. For example, it was recently proved possible to combine sublimation with HT-CVD to achieve high-quality material with a high growth rate. Also, the Smart Cut technology developed by French company Soitec makes it possible to transfer a thin SiC layer from a high-quality wafer to a low-quality SiC substrate. The final material is a bulk conductive or semi-insulating SiC wafer.
Typical SiC growth rates range from 100 µm/h to values in the order of mm/h, depending on the technique. Sublimation or HT-CVD processes provide ingots of about 20 mm in length, and each ingot is expected to give 5-20 SiC wafers depending on the polytype and orientation.
Market issues for SiC wafersAlthough larger wafers have been demonstrated, SiC material is currently only commercially available in diameters of up to 3 inches. Moreover, it suffers from defects that limit the range of applications for components, as shown in table 2.
Micropipes are larger defects that look like cracks in the bulk material and limit the available active area of the wafer. High-power-density components require large micropipe-free areas, but current micropipe densities only allow active areas of about 20 mm2.
According to numerous component manufacturers, the compatibility of 4 inch wafers with standard processing equipment makes the realization of this wafer diameter a key issue if the SiC market is to grow substantially.
With an estimated 250,000 SiC wafers (mainly 2 and 3 inch) produced in 2003, the SiC wafer market is at the start of a steep growth curve. We estimate that the volume of SiC wafers will reach more than 600,000 units in 2007.
At present, SiC wafer production is dominated by Cree, which has about 85% of the market. However, challengers such as Okmetic, Hoya and Soitec are using new growth techniques, and are expected to make inroads. Due to Cree, Dow Corning (formerly Sterling), II-VI Inc (formerly Litton Airtron) and several smaller companies, around 94% of SiC wafer production is carried out in the US, with 4% in Asia and 2% in Europe.
Until now, 90-95% of SiC wafers have been consumed in optoelectronics applications, where they are used as substrates for blue and green LEDs and violet laser diodes (figure 1). However, other substrates, in particular sapphire and also silicon or even free-standing or bulk GaN, have been used to grow nitride-based epitaxial structures.
Nitride devices that use SiC as a substrate (as fabricated by Cree and Osram Opto Semiconductors) represent about 25% of the total nitride optoelectronics market. Since no new players are entering the market using SiC substrates, we expect the SiC market share in optoelectronic applications to remain the same or decrease.
Power electronicsThe challenge for SiC is to move from the optoelectronics market to power electronics, where components can benefit from the material properties of SiC, resulting in high power densities and high-temperature operation.
Depending on the power range, many SiC components can be manufactured using semiconducting or semi-insulating material (mainly 4H or 3C polytypes; figure 2). The first commercial SiC components were Schottky diodes and MESFET RF transistors.
SiC Schottky diodes are the most advanced components using SiC heterostructures as the active layers. These components are currently in production at an estimated level of 5 million units in 2002. This is a very low figure compared with the production of standard silicon Schottky diodes; one single European producer, STM, manufactured 220 million silicon Schottky diodes in 2002.
The only two producers of SiC Schottky diodes are Infineon (through its SiCed R&D lab) and Cree, in collaboration with Microsemi. A number of other companies including Fuji, Hitachi, International Rectifier, Matsushita, STM and Toshiba have strong R&D programs related to SiC Schottky diodes, but have not commenced production.
Schottky diodes target the high-speed, high-power-density switching market. This includes products or functions like high-frequency power supplies, power factor correction and power conversion in motor controls or power management appliances. The use of SiC components will dramatically reduce the number and the size of all passive components in such systems.
Today, SiC Schottky diodes are mainly targeting the market for components operating at 6-8 A and 600 V, for power factor correction in high-end AC/DC power supplies and uninterruptible power supplies (UPSs). The total accessible market for this segment is estimated to be 200 million parts per year.
The next promising market for SiC diodes is low-end power supplies for use with PCs and home appliances. These markets will be accessible only when the average selling price (ASP) of the SiC parts falls below $1.
Technical and financial bottlenecks are still slowing down the emergence of commercial products. The SiC material accounts for about 40% of the cost of a typical SiC Schottky diode, a further 50% is processing costs, and the remaining 10% is packaging and test. Due to the cost of materials and the need for highly specialized equipment and processes, components manufacturers are hesitating to launch large-scale development in this field.
On the other hand, potential end users such as manufacturers of power supplies or UPS systems, communication base station providers, and rail traction and automotive manufacturers are waiting for new and improved solutions to manage high-power-density systems. However, implementation of SiC components is not a chip-to-chip replacement and requires a redesign of the electronic circuits.
In the RF/microwave field, SiC MESFETs are competing with GaN HEMTs to replace incumbent silicon LDMOS technology in base stations for mobile phones and broadcast (UHF band) communications. High-frequency components are also expected to be used in the L and S bands for military applications such as radar systems and jamming. Microwave heating could become a huge market if the ASPs of SiC MESFETs become compatible with the cost of the currently used magnetron technology.
SiC MESFETs have entered a pre-production level at Cree, Thales and others, but lifetime and production yield remain key issues when moving to full production.
Conclusions and future trendsSiC material offers a wide range of characteristics, with three polytypes covering the main applications. Three inch semiconducting SiC and 2 inch semi-insulating wafers are now available, but the real breakthrough will be the emergence of 4 inch wafers that are compatible with standard semiconductor tool sets.
There are three production techniques for growing SiC (sublimation, HT-CVD and heteroepitaxy), but micropipe density remains too important to manufacture numerous large devices (>2 kV) on a single wafer.
The SiC material market certainly suffers due to the Cree monopoly, with the company acting as a reference for quality and price level. No clear challengers are emerging, but the market will benefit from having a robust second source.
In terms of markets, optoelectronic applications are the leading consumer of SiC material, and consumption is forecast to increase by more than 20% per year. On the other hand, blue LEDs are now available on silicon (although brightness and yield must be improved), the first violet laser diodes on the market use sapphire substrates, and major R&D efforts are focusing on the development of free-standing GaN. This situation means that SiC will lose market share in the future, even if it shows the best lattice mismatch with nitrides and allows the growth of devices with back-face contacts. The market for optoelectronic devices grown on SiC is forecast to reach $500 million in 2003 and $1.2 billion in 2007, a compound annual growth rate (CAGR) of 24% (figure 3).
The alternative market is power electronics, where SiC enables a high switching capacity at high frequency, with no recovery time, compared with silicon. Only Schottky diodes have reached commercial production to date. The cost of these components allows SiC Schottky diodes to target high-end markets with volumes of about 200 million units per year. By reducing ASPs, Schottky diodes will be able to target the PC power-supply market and low-end, cost-driven applications. We consider that the 2003 market for power devices is worth about $60 million, and will grow to around $500 million in 2007, at a CAGR of 70%.
The market is waiting for full SiC switching cells (SiC diodes and SiC transistors) for applications such as 42 V automotive systems, but SiC MOSFETs are not yet ready, and only hybrid solutions (SiC diodes and silicon transistors) are emerging.
For high-power (>2 kV) devices, pin diodes and BJT or GTO devices are not compatible with high micropipe-density wafers and the quality of thick (>100 µm) SiC epitaxial layers has to be improved.
In the RF field, a single SiC (or GaN) transistor is supposed to replace 10 standard transistors (mainly silicon LDMOS or GaAs transistors) as used in today s base stations. This market represents a potential for 1 million components per year including defense applications. The next challenge will be the replacement of magnetrons in microwave ovens. The market for RF transistors using SiC or GaN is expected to reach $400 million by 2007.
However, SiC components do not offer chip-for-chip replacement, and systems have to be redesigned in order to accept SiC components. This constitutes the main challenge for SiC component manufacturers, who will have to teach their own customers (the end users) how to implement this new component in their systems.
