A success case of interaction between academia and industry: the silicon carbide production in Catania
Francesco La Via
The interaction between academia and industry has always been problematic in Italy where traditionally a large separation between these two actors of the innovation has been observed. In fact, from one side, the academia has always considered all the industrial problems as of low scientific impact. On the other side, the industrial researchers consider the research topics of the academia too far from the actual technology and from the market. Furthermore, the small dimension of Italian companies is generally an obstacle for the development of a long-term research in collaboration with academia. With this large gap between academia and industry their interaction has always been difficult, and only a few cases of good collaborations are present in our country.
In this paper a success case of interaction between academia and industry will be reported and the advantage gained from both sides will be explained in detail by analyzing the research activity that has been performed in Catania on silicon carbide both from a point of view of the material and on the processing and power device aspects.
2 The reason for the choice of silicon carbide for power devices
Emerging wide band gap (WBG) semiconductor devices based on both silicon carbide (SiC) and gallium nitride (GaN) have the potential to revolutionize power electronics through faster switching speeds, lower losses, and higher blocking voltages, relative to standard silicon-based devices. Additionally, their attributes enable higher operating temperature yielding increased power density with reduced thermal management requirements.
Silicon carbide (SiC) is a material presenting different crystalline structures called polytypes. Amongst these only two hexagonal structures (4H-SiC and 6H-SiC) are commercially available and the cubic form (3C-SiC) is an emerging technology. All these materials are broadly similar with high breakdown fields (2–4 MV/cm) and a high energy band gap (2.3–3.2 eV), much higher than silicon.
Different materials can be competitive with SiC for power applications, as reported in Table I.
From this table it can be understood that SiC and GaN can work in the same range of breakdown voltage but SiC is more suitable for high power applications thanks to the high thermal conductivity, while GaN is more suitable for RF applications thanks to the high saturated electron velocity.
The relatively narrow band gap of 3C-SiC (2.3 eV) with respect to 4H-SiC (3.28 eV) is often regarded as detrimental in comparison with other polytypes but it is in fact an advantage. The lowering of the conduction band minimum brings about a reduced density of states at the SiO2/3C-SiC interface. Therefore, a Metal Oxide Semiconductor Field Effect Transistor (MOSFET) on 3C-SiC has demonstrated the highest channel mobility of above 300 cm2/(V s) ever achieved on any of the SiC polytypes, promising a remarkable reduction in the power consumption of these power switching devices.
Typical figures of merit for power devices suggest that SiC is approximately ten times better than Si in terms of device on resistance for a given operating voltage and also in power density per unit area. Today 4H-SiC is the preferred material but its main limitation is the low channel mobility of carriers, which reduces the performance of the MOSFET switch used in high power applications. This limitation is extremely important especially in the region below a breakdown voltage of 800 V where DC-DC converters and DC-AC inverters are needed for electric vehicles or hybrid cars. Currently, silicon power devices are used for these applications where the inevitable power dissipation requires the use of very heavy and expensive heat sinks to keep the device junction temperature in the range where Si devices are able to function, because of the low Si band gap. Recently few Electric Vehicles (EV) builders (TESLA in primis) have produced the first SiC electric cars with a large improvement both in the performance of the cars and in the reduction of the power dissipation. These new devices were realized in Catania thanks to this collaboration between academia and industry.
3 The history
The history of silicon carbide in Catania started in 1996 at the Physics Department when the group of Prof. Foti started the first works on the formation of silicon carbide by ion implantation, studying the effect of high energy laser impulse on this compound and the effect of ion implantation damage on the optical properties of this material (fig. 1).
This activity was explorative because the material was new and no information on the several parameters was reported in the literature. Furthermore, the first SiC wafers were grown with a 1 inch diameter, the density of defects was high and the cost was high too. Few years later the activity on this material started also inside the CNR-IMM with the collaboration between Dr. Raineri, Dr. La Via and later with Dr. Roccaforte. The activity in CNR-IMM was focused on the realization of the first devices, therefore several studies on the oxidation, ion implantation and metallization contacts were performed in those years. The first Schottky diode with good electrical characteristics was produced in the IMM clean room in 2002 and in those years the interest of STMicroelectronics started to increase.
In fig. 2 the first I-V characteristics of Schottky diodes with ideal behaviour in a wide range of temperatures and currents are reported. It is possible to observe that the current vs. voltage characteristics present an exponential behavior over 11 decades and in a large range of temperatures. These Schottky diodes can carry large currents with a breakdown voltage of over 600 V and for this reason are extremely interesting for applications in several market sectors (inverters, power supply, etc.).
The main obstacle for the industrialization of this new device was the size of the substrate. In fact, it was difficult to process the two-inch SiC wafers available at that time in the standard 6-inch production line and several adaptors needed to be realized to process these small wafers. In 2004 the first Schottky demonstrator entirely realized in STMicroelectronics clean room was obtained and in 2007 the production of Schottky diodes started on 3-inch wafers. Two years later in 2009 the first MOSFET demonstrator was obtained, and the production started in 2014 on 4-inch wafers. In this period different generations of Schottky diodes have been produced and all the steps for the optimization of the MOSFET process have been developed in collaboration between CNR-IMM and STMicroelectronics. Three years later the second generation of the MOSFET device was produced on the new 6-inch wafers and then the production volume started to increase considerably.
In the Milan area in 1972 a small-medium enterprise called LPE was founded to design, produce and commercialize epitaxial reactors for silicon. Since the beginning, LPE was interested and involved in the polycrystalline deposition of SiC. For this reason in 2001 LPE decided to start a project to build a furnace to grow silicon carbide ingots using the High Temperature Chemical Vapor Deposition technique.
In the same period as the collaboration with STMicroelectronics, both the University and CNR-IMM have started a collaboration with LPE to execute a project finalized to the development of a new epitaxial reactor for silicon carbide. After some preliminary studies it was decided to realize a hot-wall reactor that gives the best material quality and the highest growth rate. The chamber of this type of reactor is schematized in fig. 3. The whole chamber of this type of reactor is in graphite and it is heated at high temperatures (1500–1700 °C) by several coils that induce a high current in the graphite walls. The temperature in this way is very uniform because all the walls irradiate the heat, and no temperature gradient is observed inside the chamber. The only non-uniformity in the temperature field is introduced by the gases that have to be introduced in the chamber through the inlet to produce the reaction and the deposition on the substrate. To solve this problem and also that of the natural consumption of the precursors, the wafers need to be rotated to increase the thickness and doping uniformity.
The first prototype of the reactor designed and produced by LPE was installed in STMicroelectronics in 2003 (fig. 4). In this way a new activity on the development of the epitaxial growth process started in the academia and several new reactors were realized by LPE to improve both the throughput and the epitaxy quality.
Between 2004 and 2007 a lot of studies on the development of the process were performed and a new process with the introduction of HCl at the beginning and of trichlorosilane (TCS) later produced a large increase of the growth rate and a new era in the epitaxial process of silicon carbide started. As reported in fig. 5, the growth rate increased with the new process from 6–8 $\mu$m/h to 120 $\mu$m/h. This new approach opened the possibility to the realize a new class of devices with breakdown voltage of 10 kV and new detectors for nuclear particles and neutrons.
The possibility to grow at high growth rate produced not only an increase of the throughput of the reactors but also an improvement of the material quality in terms of better surface and lower defect density on the wafers. In fig. 6 three different photo-luminescence maps of three different wafers grown at different growth rates are reported. From these measurements it is possible to observe that the density of stacking faults (one of the main crystallographic defects of 4H-SiC) is considerably reduced increasing the growth rate. This kind of defect is considered a killer defect for several devices and then this new process with chloride addition produced a considerable increase of the yield of the devices and subsequently a reduction of the costs.
In 2006 the first commercial reactor ACISM8 was produced and sold to different companies and universities around the world. The reactor can grow 6 two-inch wafers or 3 three-inch or 1 four-inch ones. In order to increase the throughput, a new reactor having a bigger process chamber and higher throughput (ACISM10) was designed and produced in 2008. The throughput of the epitaxial reactors was the main bottle neck of the SiC technology and then in the following years a large effort in LPE and at CNR-IMM was devoted to the development and the realization of a new reactor with a higher throughput. LPE decided to realize a single-wafer reactor for 6 inches that could be easily scaled up to larger diameters and obtain a higher production throughput by decreasing the process time. To reach this goal, a loading chamber was realized and the wafer was introduced at high temperature in the growth chamber. In this way the process time for the temperature ramps (up and down) and the evacuation of the chamber was greatly reduced and a typical growth process time was reduced from four hours to one hour. With this new reactor a large improvement of the fluid dynamics of the precursors has been performed producing also an increase of the thickness and doping uniformity. After this big step in the improvement of the reactor design, a further problem to solve was the wafer loading automation. The SiC wafers are transparent and then it was not easy to realize this automation using standard technology, but after several studies and experiments the new cassette-to-cassette automatic reactor was realized in 2019, as a part of a joint project among LPE, STM and CNR-IMM.
In the last two years, the device companies have requested the market to use silicon carbide wafers at bigger diameters: 200 mm. For this reason, ST, LPE, and CNR-IMM had again joined resources to develop a 200 mm epitaxial reactor as a part of a big EU project to set up a 200 mm silicon carbide device pilot line. The 200 mm epitaxial reactor was delivered to ST in March 2021.
4 Impact on the academia and on the companies
This large effort on both the development of power devices and the development of materials has produced a considerable impact on the academia both in terms of publications and of new projects. If we look at the distribution of the publications in the last 20 years (fig. 7a) when these collaborations on the silicon carbide started, we can observe a large increase of the CNR-IMM publications on silicon carbide and of the citations/year. The same behavior is observed for all the Catania collaborations (CNR-IMM, University, STMicroelectronics and LPE) for both papers and citations/year (fig. 7b).
These publications can be used to follow the main topics of the investigations in these years. In the beginning of the SiC study essentially the amorphous silicon carbide has been characterized and studied by optical measurements. Subsequently, the study has been moved to crystalline material and to the study of the main metal contacts: titanium and nickel silicide. For several years the activity has been focused on the realization and characterization of Schottky and Ohmic contacts. Starting from 2004 a new activity on epitaxial growth was started and a new process with chloride addition was proposed. In 2006 also a new Schottky UV photodiode was proposed and the year later the first paper on the 3C-SiC growth on silicon appeared. In 2008 the first work on epitaxy using trichlorosilane (TCS) appeared in Journal of Crystal Growth and it was demonstrated that a high growth rate (100 μm/h) can be obtained both with the addition of HCl and with this new precursor. In 2009 the first paper on Schottky diodes on 3C-SiC was published and the year later the first 6-inch 3C-SiC wafer was shown. Two years later, in 2011, the first works on SiC MOSFET and the first papers on MEMS were published. In the last years several review papers have been published on dielectrics, epitaxial growth, technology for power devices and 3C-SiC growth.
This collaboration has been supported by public grant and for IMM in the last years a total amount of about 8.5 M€ with different national, European and industrial projects has been obtained. Also, the University participated in different projects with a total grant of about 6.2 M€. In the case of LPE the grant from the research projects in the last 20 years was of about 13.7 M€, while in the case of STM it was of about 27.3 M€ (fig. 8). It is interesting to note that the SiC activity started before in IMM, LPE and the University essentially on the development of the material and later in STM on the development of the devices. Among these projects I would like to recall the SiCiLab project that was devoted to the realization of a new private/public laboratory that works on the development of new epitaxial reactors, new epitaxial process and new characterization techniques for SiC wafers. In this laboratory a high growth rate process, a new large reactor (ACISM10) and a new bulk reactor were developed. Furthermore, several characterization techniques (XRD, µ-Raman, µPL) have been implemented for SiC characterization.
Another large project that has a large impact on IMM is CHALLENGE. In this project, coordinated by IMM, several CNR groups have participated in developing a new silicon carbide polytype: 3C-SiC. This polytype has a smaller band gap with respect to the most used 4H-SiC and it is particularly interesting for the applications in the electrical vehicle technology. Also, in this case the academia is going towards a new way and, if the project succeeds in developing the material industry can have a large benefit due to new products with better performance than the traditional ones.
In this project STMicroelectronics, CNR-IMM and LPE have developed a new process and a new reactor design for the realization of a bulk 3C-SiC wafer. This revolutionary process is reported in fig. 9. After the standard hetero-epitaxial process of growth of the 3C-SiC layer on silicon, with a carbonization step at low temperature for the formation of a seed, and a further growth step at higher temperature, the temperature in the reactor is further increased above the melting temperature of silicon and the substrate was removed inside the reactor. Then the temperature is further increased to increase the thickness and obtain a 3C-SiC wafer. In this way both 100 mm and 150 mm wafers have been obtained. The advantage of this process is that it is possible to obtain a good material on large wafers without the long and expensive procedure of enlarging the ingot as in the case of 4H-SiC.
This 3C-SiC wafers have been deeply analyzed with several techniques and the crystallographic defects and their interactions have been studied in detail to understand their behavior and the interactions between different defects (stacking faults, anti-phase boundary, point defects, etc.). In fig. 10 several different types of Stacking Faults (SFs) both in atomic resolution and in low resolution are reported. Three different types of stacking faults can be observed. The first one is the classical SF, where a single plane is rotated by 180° along the (111) plane. The second one is a SF with a rotation of two layers and the third one is a rotation of three layers. These different SFs have different formation energies and different behaviors during growth and can be influenced in different ways by the process. These SFs are limited laterally from different types of partial dislocations and these dislocations can have a large impact on the evolution of these defects during growth. Finally, during the CHALLENGE project, it has been observed that these defects can produce a leakage in the devices and then a large effort has been devoted to the reduction of these defects to obtain better device performances.
In the last few years also two large ECSEL-JU European projects have started in Catania. In the first one, the WInSiC4AP project, several technology bricks for efficient and cost-effective applications addressing social challenges and market segments, in which Europe is a recognized global leader (e.g., automotive, avionics, railway and defense) have been developed. In particular gate oxides, contacts, passive components, interconnection technologies, etc., reliability and failure modeling analyses, advanced packaging and bonding options, up to the design methodologies for high efficiency applications, have been studied in detail. The second one is the REACTION project, finalized to the realization of the first worldwide 200 mm silicon carbide (SiC) pilot line facility for power technology in Catania. This will enable the European industry to set the world reference of innovative and competitive solutions for critical societal challenges, like energy saving and CO2 reduction as well as sustainable environment through electric mobility and industrial power efficiency. By establishing the first 200 mm SiC pilot line in the world and developing the most innovative and cost-competitive technology, this project will address mass-market applications like smart energy and smart mobility, as well as industrial applications. It will allow to meet the more and more increasing demand of requirements in terms of quality and cost constraints for next decade generation’s power electronics.
Another project that can open a new market for silicon carbide is SiC nano for picoGeo. In this project, coordinated by IMM, 3C-SiC is used to realize a new MEMS of high performance for geophysical applications. In this case the outstanding mechanical properties of 3C-SiC (high Young modulus and high Q factor) can be used to realize a very sensitive strain-meter that can measure the deformations of volcanos in the 10–12 range. With this sensibility it will be possible to see the small deformations before eruptions and earthquakes, solving one of the main problems of geophysics: eruption and earth wake forecasts.
Also, in this case the academia is exploring a new field of application that can be used by several companies in a near future to develop new business in new fields.
This collaboration led also to the organization of different conferences and workshops as reported in Table II. Starting with the HeteroSiC-WASMPE in 2009, to the ICSCRM 2015 in Giardini Naxos that is the major conference on silicon carbide with a large participation of scientists and companies coming from all the continents. In the last years two workshops connected to different European projects have been organized too (First International WInSiC4AP Workshop, Silicon Carbide in Europe 2020) and a symposium at the E-MRS, one of the largest European conferences.
This activity has produced also a large number of Master and PhD Theses among the University, CNR-IMM, LPE and STM especially in the first period of collaboration when several fundamental studies have been done (fig. 11).
In addition, the numbers of patent have increased for both academia and the companies involved in the collaboration, as reported in fig. 12. In the case of LPE the peak of the patents was in 2007-2008 due to the work for the design of the new P106 reactor and of the HTCVD reactor for bulk growth. In the case of CNR-IMM, the main activity on patents was related to Schottky diodes, MOSFET, hetero-epitaxial growth of 3C-SiC/Si and oxide isolations. Finally, in the case of STMicroelectronics the main patents have been done both on Schottky, MOSFET, UV detectors and MEMS.
This large activity has been favored by the close proximity of CNR-IMM, STMicroelectronics and LPE in Catania. As reported in fig. 13, the main part of the CNR-IMM laboratory is inside the STMicroelectronics fab in Catania since the beginning of its history in 1990. Starting from 2007 also another laboratory was built inside the LPE building in Catania. Subsequently CNR-IMM has also bought a part of the LPE site and the new laboratory and offices will be realized inside this location within the Beyond Nano Upgrade project (40 M€).
The large investments in the field of silicon carbide and the large know-how produced by the collaboration with the CNR-IMM has produced in 2018 a large increase of the power device production of STMicroelectronics and now this company is the first producer in the field of silicon carbide, as shown in the fig. 14 taken from the YOLE report of 2019. Moreover in 2019 the acquisition of NORSTEL has been done by STMicroelectronics to have an internal supply of silicon carbide wafers. In fact, European power devices manufactures depend strongly on the US suppliers for SiC substrates and in the last years it has been observed that the supply of substrates is the main limiting step for the increase of the production of silicon carbide power devices. Then all the main power devices manufactures, as STM, ROHM, INFINEON, have acquired or have attempted to acquire the few companies that produce SiC substrates, trying to considerably increase the production of SiC substrates outside the US.
For the epitaxial reactor market, the situation is more fluid, because several producers are present and, especially in the last years, there is not a predominant supplier. In years to come, the increase of wafers dimensions could also change the actual distribution of the market and probably we will see a transition from the batch reactors to the single-wafer ones. LPE has been one of the pioneers of the SiC single-wafer reactor and this fact could be a large advantage in the future market.
For LPE the SiC business has become more and more profitable during the years. The projection of the revenue for the next years is shown in fig. 15. It can be observed that the production of industrial reactors for power devices (for 200 and 150 mm wafers) will increase considerably in the future, while the production of reactors for universities and research centers (100 mm wafers) will decrease.
5 Summary and outlook
A success case of interaction between academia and industry has been reported in this paper. This interaction does not only produce an increase of the revenue of the industry but also a large increase of the know how on the processes for epitaxial growth and device processing and this can be observed by the large number of papers and patents on this subject. Furthermore, this activity has produced also an increase of the international relevance of the main actors involved and then several large conferences and workshop have been organized in the last years. Another consequence of the international relevance of the researchers involved in the silicon carbide activity is that in the last years 4 large European projects have been coordinated by a Catania researcher.
In the near future this activity will be further improved and the increase of the research employment at academia and industrial level could be expected if investment along the value chain are supported at national and European level. The CEO of STMicrolectronics in a recent interview to La Repubblica has reported that the EV market will reach 3.7 billions of dollars in 2025 and STM wishes to obtain about 30% of this market. STMicroelectronics wishes also to realize in Catania a pole of excellence in this field thanks to the collaboration between CNR and the University.
Furthermore, new applications have been explored in the last years like MEMS and detectors and new start-ups will be established in this so-called “Silicon Carbide Valley” to try to further increase the level of high-tech employment and revenues due to this material in the ecosystem.
Also new research institutes present in Catania are interested in this activity. The National Institute for Geophysics and Volcanology (INGV) is particularly interested in the realization of very sensitive strain meters to study the strain field in volcanos or in the faults. This activity, reported also in the European project SiC nano for picoGeo, will have a large impact on the geophysics applications and on the MEMS market in general, because cubic silicon carbide (3C-SiC) can have a large advantage with respect to the presently used materials thanks to its good mechanical properties.
The National Institute of Nuclear Physics is instead interested in the use of this material as a detector for nuclear particles (ions, X-ray, neutrons, etc.) both for the experiments in nuclear physics and on the applications of these radiations in nuclear medicine. In fact, silicon carbide has a lower noise level and a higher radiation hardness that are extremely interesting characteristics for these applications.
I wish to acknowledge several people that have made a considerable contribution in the SiC development in Catania. First of all, G. Foti and V. Raineri who have started this activity at the University and in CNR-IMM and that unfortunately passed away in the last years. Then, F. Roccaforte who has made a large contribution to the study of metal contacts, gate oxidation and in general device processing. In STMicroelectronics the first researchers that believed in SiC and supported this activity were F. Frisina and M. G. Saggio. In LPE I wish to thank G. Abbondanza, F. Preti, D. Crippa and M. Mauceri for their important contribution in the realization of a new class of epitaxial reactors and of new high growth rate processes. In CNR-IMM a large number of researchers have done considerable work on this subject starting from F. Giannazzo and P. Fiorenza with their contribution on the AFM electrical measurements (I-V, C-V), M. Zimbone and V. Scuderi with their contribution on µ-Raman and TEM analysis, A. La Magna, G. Fisicaro, I. Derenzis with their contributions on Monte Carlo and molecular dynamics simulations, A. Alberti with her contribution on XRD measurements, S. Privitera with her contribution on electrical-optical measurements, S. Boninelli and C. Bongiorno for TEM analysis and S. Di Franco for device realization. At the Physics Department of the Catania University important contributions to this subject have been given by L. Calcagno with her activity on the detectors and R. Reitano and P. Musumeci with their activity on the optical characterization of SiC. I also want to acknowledge my older post-docs A. Severino and R. Anzalone that have largely contributed to the 3C-SiC activity, N. Piluso that has started the µ-Raman and µ-PL characterization of both 3C-SiC and 4H-SiC and M. Camarda that started his activity on the Monte Carlo simulation of silicon carbide epitaxial growth and now is the owner of a start-up that is working on SiC detectors.